This page is not created by, affiliated with, or supported by Slack Technologies, Inc.
I tried to come up with an "in soviet russia, looking forward to talk you!" inversion and sorta failed tl;dr looking forward to yours 😄
Super excited for the Scenius to form and start the firehose of information and engagement!!
This is where all the plenary discussion will happen live
Your colleagues don't have to miss out! We're livestreaming all the sessions from ETLS Las Vegas to everyone with an upgraded IT Revolution Membership (http://itrevolution.com/membership/).
1@alexb Any issues with the livestream? Trying to get it to go but it looks like it's just at a loading screen.
So excited to welcome everyone near and far!
Livestream seems not to be connecting 😢 anyone else having success (or similar failure mode)?
You can read the paper The Checkbox Project here: https://itrevolution.com/product/the-checkbox-project/
DevOps: The Operating System for the Modern Enterprise, not just CICD-Pipelines!
People on site... please post pictures of the event on Slack if you get a chance, for that exciting real-time immersive experience 🙂
Thanks! With the livestream not working yet, you are our only link to the live lovestream 🙂
This really is a special conference. Each year has built on the last with so many industry learnings. ❤️
So sad to miss this year but wishing everyone in Vegas an amazing ETLS experience!!
Have already seen @willardamyl - would be good to have you hear Matt. Hope all is good
@nickeggleston trying to. No luck so far.
@alexb to the rescue!! Livestream is up and working!
Hi, I'm speaking today at 1655 in Azure about https://etlslasvegas2024.sched.com/event/c8ae280384d4089f52de987cec35de43 🎛️ 🙀. Here's an article on https://www.equalexperts.com/blog/our-thinking/escape-ticketing-hell-with-a-shift-to-self-service-platforms/. Let me know if you like it! 👂
Hi, I'm speaking today at 1655 in Azure about https://etlslasvegas2024.sched.com/event/c8ae280384d4089f52de987cec35de43 🎛️ 🙀. Here's an article on https://www.equalexperts.com/blog/our-thinking/escape-ticketing-hell-with-a-shift-to-self-service-platforms/. Let me know if you like it! 👂
@steve.smith - You had it right - Azure! Unless we’re having Margaritas, then Azul…🍹
Super excited for my first time at this conference and oh hey hi Jeff! Way to represent Excella
✈️ Please welcome our first speaker, Lauren Woods, Senior Vice President & Chief Information Officer, Southwest Airlines, here to present A CIO's Journey: Turning Turbulence into Smooth Skies
Not typical business application: buying airplanes ($150-300MM average order price?)
Probably not on auto-replenishment ordering
Hope Lauren shows a photo of the hosted Commodore 64 that bypassed the outage in July. 🤞
"No one cares more than people with Southwest badge" — helped set tech direction forever afterwards. (Matches what we saw in so many organizations, who shifted from 80% outsourced IT to 80%+ insourced IT — Target 2013)
So excited that Lauren is sharing details of how they responded to 2022 Winter Storm Elliott, where winter storms caused massive cancellations across the US. Dr. Steve Spear and I wrote about this in the Amplification chapter in Wiring the Winning Organization. https://www.cnbc.com/2022/12/24/winter-storms-force-airlines-to-cancel-10000-flights.html
Both the Southwest (edit: storm outage) & the Delta issues stemming out of Crowdstrike are sobering reminders that these are complex dynamic systems that can slip into unexpected states.
Wasn’t southwest largely unaffected by crowdstrike?
in both cases there was a triggering event that resulted in unexpectedly severe impacts.
Interestingly, both Southwest and Hawaiian Airlines, both who spoke yesterday, had minimal impact due to CrowdStrike.
Issues: business process problem, aircraft network, crew network; all led to situation that systems were never designed for.
Implemented self-service customer refunds (in days). We had to show them better. Created lost luggage system, at scale never imaged before (in days)
I'm curious how much these air travel metaphors are used in internal communications to frame changes.
We really can't help ourselves at the airlines. My main product is Runway. Our pipeline is the Taxiway!
We do that a LOT. Putting things in airline metaphors really helps the message to resonate with our People.
@jim.dayton You’re lucky that you have such a rich domain to draw from here, actually. I spend a LOT of my time trying to come up with metaphors to help get ideas across. Fun tmes!
"Everyone ran towards the fire" : SWA left their families during holidays to help their work families, to help our customers
12x data processing faster; developed faster than ever; Result: SWA has lowest cancellation rate of any airline!
The biggest cultural change: "we recently announced changes to cabin design: for first time, assigning seats"
"Assigning seats, premium seats... all significant. None of our apps have any concept of seats, selling seats, etc..' (!!)
Assigned seats on Southwest…reaction from my friends with young kids? Thank God! 😅
One thing I wish I had known before I started thread 🧵
I wish I had known how much of the work of software delivery and effective infrastructure is sociocultural! There are so many people / connectivity issues and if you don’t have an effective process technology solutions can often just increase the frequency of your failure conditions
Immediate thought: digital queues to board rather than have everyone cluster near the signs.
✨ Let's say hello to Michael Nygard, VP, Data Engineering, Nubank, here to present: Data - The Land DevOps Forgot
Mike's awesome book, Release It! https://pragprog.com/titles/mnee2/release-it-second-edition/
Is Mike on the list of book signings? I’m hoping to pick up some physical copies of these books I haven’t read
@joshua.andrew.phillip he is not one of the authors signing books tomorrow evening. but if you got a copy of his book, i bet he would sign it for you 😉
maybe IT Rev will have him write a book on data devops soon hahaha a girl can dream....then we can have a Nygard book signing at ETLS 💕
Hoping Space Unicorn makes a comeback this time (nudge @mvk842)
Space Unicorn was very special for the launch of “The Unicorn Project.” But maybe…stay tuned!
Coincidentally, I recently made a Grainger's Song 2024 with Suno AI using old company song lyrics from the 1930s... https://suno.com/song/d2ea15c5-14c2-4128-8271-b08955a415fe
Interestingly, previously, @mtnygard was SVP of Platform Engineering at SABRE, a key part of the travel/hospitality industry. Later today, we have an amazing talk from CLEAR and Hawaiian Airlines. Today is travel industry day!
Nubank: 50 PB data ingested nightly, 180K+ jobs, hours run time, deploy time in hours. The new danger: merge queues!!! 😱
A daily Analytic job that runs more than 24 hours is a bit of a problem
Those 8 hour build and test failures must have been so costly when they inevitably ran into issues with their CI runners
Absolutely. A huge source of delay, cost, and frustration. It got into "one change per week per data scientist" territory
Making product owners responsible for the quality (and quantity) of the data they generate! (E.g., the length/cost of the batch jobs that depend on them!)
Love it. Also love costs being attributed back to the PO as an important feedback loop for understanding value.
4 Key Principles • Domain ownership • Data as a product • Self-service platform • Federated governance Will be using these as a litmus test later…
New choice: make files and data formats private: expose only the data and schema!
I love that this conference has talks on batch jobs — there is another fantastic talk on batch jobs on Day 3 from the Vanguard team! 🎉
Slack is the ultimate (but not ideal) data discovery mechanism. 😂
Select * from enterprise_data_whs. Run the algo. Export to excel
We’ve had some high stakes data projects where we needed lots of independent decision making occurring asynchronously in a short period of time (imagine a hackathon with millions of dollars on the line) and we used a similar concept called Data contracts or data promises between teams to maintain apis or informal api specs or data structures as the cardinal step in our process
There is ANOTHER great experience report from Vanguard on creating internal data marketplace.
Changing an OR in data query reduced job from 900m to.. 1m. (Did I get that right?)
And that's just one example which is recent enough for me to remember the details.
That’s what I’m hearing :). I have these roles in my organization and this talk is validating the decisioning here (also giving me some 💡 moments on what I missed)
AI is going to have significant impact in helping find optimizations in inefficient queries
"Even orgs that are only 10 years old have legacy". 😂
Code bases 3 months old often have legacy! At least if the org has high levels of context switching and reactive processes
And nothing beats the fun of blocking and locking in SQL queries
Just like in prompt engineering the questions we ask (queries) our data is critically important.
By the time you're sick of saying something, they just began hearing it.
After 3 presentations to the org / CTO and innumerable async updates on our roadmap and current progress in under a month this one hurts that it’s true
"if people don't hear something in six weeks, they'll assume it's dead / been killed." 😂
"By the time you're sick of saying it, they've just begun to hear it"
That advice applies to parenting as well. I can speak from experience as a proud Dad of 3 kids.
Feedback loops: cost attribution! One group that owns all the data job costs, that anyone can dump more work into — of course, costs go up uncontrollably! 🔥
Feedback loops: cost attribution! One group that owns all the data job costs, that anyone can dump more work into — of course, costs go up uncontrollably! 🔥
Hard to do .... IT Finance still struggles with "dollars chasing dimes." FinOps has helped it progress...
This is true for many layers within the tech stack. Most scaled IT orgs have cost accounting at the bottom of the priority list in good times and bad. Bright spot - I know how to solve this and have done so in several organizations! Ping me if anyone needs help here.
Another platform group in Nubank created a Cloud Financial Management team about two years ago. I think they've paid for themselves for the next 25 years already.
"If you say something once and wait 6 weeks they will assume it's dead because they have not heard about it in 6 weeks. " Love that.
I’d like to subscribe that any answers to @mtnygard ‘s requests at the end. :)
The thing I liked around this comment was to not leave the legacy world behind, and to create the space/improvements now to allow everyone to contribute to the new world.
🎸 Please welcome the team from SiriusXM – Ben Manning, VP Services Engineering; Rachel Uhrig, VP of Ecosystem Reliability; and Carol Gastal, Lead Strategy Partner, Equal Experts. They're here to present How SiriusXM Built a Next-Gen Streaming Platform for 34 Million Subscribers in Less Than a Year
That was sooo good @mtnygard Well done 👏
had a sense of full circle ... I built Target's metadata repository (data catalog) in 2000 for data discovery... it was hard to explain to some folks why we were even doing this -
Wow! We need to talk about the then and now @char
See https://en.wikipedia.org/wiki/Musical_anhedonia. "Musical https://en.wikipedia.org/wiki/Anhedonia is a neurological condition characterized by an inability to derive pleasure from https://en.wikipedia.org/wiki/Music."
Thank you for sharing this. I learned something new today.
So, as it happens, Larry Page & Sergey Brin hated music back when I was at Google. They didn’t listen to it and didn’t care about it. Larry had one CD and it was a gift from Bono. I know this because we did a surprise Music Trends project (sort of an early Spotify) and they loathed it and told us never to work on it again. 😆
Transition to "satellite broadcast company" to ____, not necessarily tied to hardware device in the car. (Can someone fill in that blank?)
"Tightly coupled, loosely controlled" — worst of all worlds.
"CEO announced we're going live, months before we got there" — aggressive timeline, could not fail.
Yes it created a strong alignment point, which everyone could relate to 👍
Why do so many of these shared experiences give me flashbacks? :melting_face:
Stream aligned teams created autonomy, all heading towards common goals
We are currently implementing this at Brivo (after reading Wiring the Winning Organization)! There have been grumbles by the old guard but I’ve seen huge improvements to our ability to roll out horizontal slices of process improvement. Huge impact to psychological safety, reliability, and reductions in major outrages
Architecture Decisions documents decisions so that other teams can easily access and read (and understand) choices made.
Challenge: how do you enable teams to make decisions quickly, independent, at speed. Created _ meetings facilitate quick decisions, and maintain coherence.
stream-aligned teams still need to ensure they aligned with the overarching goals and provide high communication points. well said @carol.gastal
Yep it's been a huge part of the SXM journey @david.faircloth - we should all swap notes
Question: what are people using for traceability in order to track these impacts? That’s my current challenge
We have datadog but lots of ad hoc dashboards at the moment. Do people have dedicated teams for this? Tools? Magic lamps?
No silver bullet, and lots of caveats - but much of the conversation is around OKRs. Unless I'm not quite grasping the topic here.
are you asking from a technology perspective or an org/culture perspective? DORA metrics is what we really dive in to to help craft the story of change and showcase easily where we are going and continued improvement.
@joshua.andrew.phillip Larger orgs may start to build out an architecture for this
@nathen.harvey is here, you should get with him as he's the Developer Advocate at Google for DORA. He's helped us a lot with our setup and approach.
1
1May be worthwhile to look into Datalake as an option
also, shameless plug to come to Capri-6 @2pm for the Wendy's/Navalia discussion around this.
1 1I’m having trouble getting buy in from C level leadership on spending serious time (3-6months) on traceability. I’m not hung up on what to track but really how to track it and how to do that “as we go” with minor bets in our iterations
We implemented a Metrics Lake (similar to a data lake). Everyone has access to all product metrics, free to rely on DORA, SPACE or whatever gives them the visibility to reach their goals. Same proccess is also available for our data lake, giving our business division the tools to make queries in our datalake.
I’d definitely love to hear everyone’s experiences! This has been a 3 year battle
We have began using OpenTelemetry to apply the practice of observability for insights and signals to the Engineering Capabilities and Practices.
@joshua.andrew.phillip - @ericc is demoing how to set up collecting metrics like DORA via OpenTelemetry at the Liatrio both. @ericc , is our open source work relevant here? Sounds like it might be.
3Yes sir @matt.sibley! Here is the link and OTEL gives you platform agnostic standard.
Gosh that metrics lake sounds like a dream @louis-alexandre.berge Any background on how the org / senior leadership came to make that big bet?
I got autocorrected!! I meant https://devlake.apache.org
I’ll check that out, I’ve followed the OTEL CI/CD SIG / convo. Definitely would love to chat more on that
I’m working with the maintainers to extend it to a fuller value stream scope
Here is the https://github.com/liatrio/tag-o11y-quick-start-manifests to the above referenced open source quick start. Create a PAT token and off you go with SCM related data.
1So many good ideas here, can all of you also call my CTO ? :troll: jk I really do appreciate it I’m hoping to leave here with a solid proposal on this so this is all gold. Looking forward to picking your brains
@steveelsewhere this is exactly where my mind went… how do we offer data to back the VSM and Flow.
@joshua.andrew.phillip Higher management speaks in terms of money. So to implement our strategy, we architected a centralized metrics engine that would not only format metrics for every team, but also save 80% of our costs along the way. We embraced an early "You build it you run it" mentality 8 years ago. That led to a lack of standardization. Our architecture not only reduced the cognitive load on teams, but also made it easier for everyone to browse through all of these metrics.
@joshua.andrew.phillip VSM: Value Stream Management
We are leading and would love more maintainers to contribute to the OTEL community. We just lead the release of Semantic Conventions with CI/CD focused attributes - https://opentelemetry.io/docs/specs/semconv/attributes-registry/cicd/ this will standardize the data coming from CI/CD systems.
@joshua.andrew.phillip if you’re interested in value stream mapping definitely come to the learning sprint with myself and Andrew Thursday! https://sched.co/1iGue
I really want to know more about “Organization wide Friday demos”
Interestingly, both Benjamin and Rachel are both Disney+ alums —
2With 12 month timeline, there wasn't room for errors and rework.,
"Beg for forgiveness" This is so underrated. Move mountains, don't ask!
Kudos for asking forgiveness not permission to visualize work on conference room glass
Kudos to overnight Graffiti’ing the CTOs conf room! It’s those types of acts and moments that can fundamentally inspire and align an entire team.
Results: during launch, zero Sev 1 or Sev 2???? No way. WOW!
⚡Up next, Michael Carr, Principal, Vanguard Chief Technology Officer, Vanguard, here to present: Technology Leadership To Help Clients and Crew at Vanguard
Gene attending the three day event produced lots of good talks for ETLS
Launched 3 strategic partnership post-launch: first was Hilton, and the rest to be announced soon. Kudos!!
$9.5T assets under management, 20K employees, 50MM investors, 10K technologists (crew and contractor)
“Modern organizational principles” 👀 Looking forward to hearing those…
8500 crew participate in our sessions — 50/50 split between technologists and business!!!!
That guy in the picture looks familiar :hmm:
Love how Gene leans-in to the conference experience as an attendee
Seriously — the Vanguard internal tech conference was amazing — so many amazing people working on amazing things. Four Vanguard teams are presenting here this week. They're amazing! 🎉
Can confirm, @genek was ridiculously excited and remains extremely enthusiastic about our Vanguard friends!
“End-to-end modernization gives us a pace advantage.” Yes! Speed wins!
@genek , you are just about to get invited to all our Conpanys’ internal events… including adidas for sure!
"We're not making money from our clients, we're making money for our clients" — Salem Ramji, CEO
"Many people selling financial products with... interesting claims". 😂
"Buy low, sell high. Greed makes you buy high, and fear makes you sell low." <--- the cognitive biases that make investing genuinely difficult.
The “spare time” investor. Is that a course I missed in college? :face_with_monocle:
If it’s anything like my college investing experience you’re lucky you missed that course 💸 :melting_face:
“We need a mutual fund industry with both vision and values. A vision of fiduciary duty and shareholder service, and values rooted in the proven principles of long-term investing and of trusteeship that demands integrity in serving our clients.” – John Bogle (founder of Vanguard)
Goal: "subtly nudge our clients towards better investment decisions through awesome CX (customer experience)" — test and learn to find the best nudges.
“Progressively better client experiences.” Sounds like a Progressive Delivery model.
5x increase in the pace of change with technology and 75% reduction in incidents. Speed AND stability.
"We had teams iterating and pivoting into oblivion for no purpose" — Matt Benchener
Followed by “Autonomy and empowerment without clarity and accountability is chaos.”
@mtnygard ‘s presentation was outstanding! Thanks for all the tips here! So valuable!
"Now some microservices are having 'moving parts' problem — fewer incidents, but half the company now has to be on the outage call". OMG. Hahaha. What a great characterization.
Microservices definitely trade coupling for cohesion(lack of). I like to tread carefully in adding them arbitrarily. Distributed systems are hard... :)
It’s all fun and games until you have multiple microservices running in PROD. Gotta up your game in observability and resiliency engineering. It’s certainly a paradigm shift.
Implementing Tracer Bullets and correlation ids to help with transaction traceability across micro services has been extremely helpful
Perhaps a channel here on slack or some other way to stay connected and share these stories, with some way to anonymize for those who need it.
Rolling change out across multiple teams and feeding back lessons learned across the org
🚜 And now, the team from John Deere, Amy Willard, IT Director, Global IT Strategy & Transformation; Justin Thomsen, Group Product Manager - Developer Experience Platforms and Adam Brunner, Principal Software Engineer and Group Product Manager for Developer Advocacy. They are here to present: John Deere's Transformation - Learnings on being Value Driven and Employee Centric
Technology can slow modernization to a crawl. Culture is the much harder problem to solve.
@willardamyl group: centralized group that helps every other technologist and knowledge worker to their best work: includes UX design system, agile & product platform, digital mastery & advocacies, quality/SRE/operational risk, AI & citizen development, DevEx her fave part: the *: partnering with security, architecture, and embedded change agents. Her summary: Central organizations are great!
"No human to machine interactions" — I.e., can't log into system to deploy.
Human having access to production —> I don’t trust your pipeline no matter what you say
Here are all the things that we expect all software people to be experts. Hahaha.
Thank you for attending our conference for the last decade @justinrobertthomsen! 💚
These presenters are 🔥 and elevate and inspire the rest of their org. Whoever gets to work with these people every day is ridiculously lucky! 🙏
Fascinating idea for DX and platform teams: have town halls for developers
I believe @jonathansmart1 calls that incentive... 
when the conversation is why are other teams kicking your asses, you know you're on the right track
IIUC, John Deere has massive Java footprint, thousands of applications, many decades old — log4shell response must have been epic. Cc @mring
Curious - have folks tried branding their DX platform with a name? I’ve seen orgs go so far as to even make tee shirts with a custom platform logo , etc to drive developer adoption.
Yes! We named our platform Runway. We have a logo, various sticker runs, custom animated slack emojis, cups, tshirts and so much more
We do, we called it adidas digital platform, and we have thousands of T-shirts with our #adidoescode logo
Yes! The name of our platform was one of the biggest asset. Have mugs, stickers (with glitter 🪄), and we aim to distribute them based on « achievements ».
I've done that before. Voice of experience: don't include "New" or "Next Generation" in your platform branding. By the time you get adoption it won't be. 😂
Internal employees don't (normally) pay for your products (i.e. increase company revenue) but being customer-centric with internal employees does help accelerate both end-customer and company outcomes.
Improved business outcomes that... • Increasing revenue (improving flow of changes from idea to release) • Protecting revenue (improved system quality, reliability, availability) • Reducing costs (through automation, improved productivity) • Avoiding costs (better in-place safety/security/compliance checks).
Source of this model: https://blackswanfarming.com/value-a-framework-for-thinking/
“Stories are the unit of idea transmission.”
Love the shout out for Investments Unlimited and how they incorporated the principles. ❤️ 🙏:skin-tone-3: :hugging_face:
Minimum definition of deployable... sounds like MinimumCD @bryan.finster486
@genek have you ever considered inviting the anti-heroes from some of these stories?
The simultaneity of breakout sessions is sooooo FOMO-inducing 😆
Need a "physical twin" who is on site attending the BOF sessions ... 🙏
Don’t Miss Cloudflare’s Demo at 12:30pm today at the Solutions Hub located in the Expo Hall. https://etlslasvegas2024.sched.com/event/1hAXl/solutions-hub-cloudflare-what-if-the-10x-developer-was-all-the-developers-you-hired-along-the-way In this live coding demo we will explore what the future can look like when you can chat with your code and generate scaffolding, fix bugs, and even pivot to new use cases. (edited) http://etlslasvegas2024.sched.com https://etlslasvegas2024.sched.com/event/1hAXl/solutions-hub-cloudflare-what-if-the-10x-developer-was-all-the-developers-you-hired-along-the-way View more about this event at Enterprise Technology Leadership Summit Las Vegas 2024
Don’t Miss Sonatype’s Demo at 1:00pm today at the Solutions Hub located in the Expo Hall. https://etlslasvegas2024.sched.com/event/1hAXH/solutions-hub-sonatype-experience-sonatype-sbom-manager Discover how SBOM Manager can streamline and automate the auditing, distribution, and monitoring of SBOMs at scale, ensuring compliance with the latest regulations. http://etlslasvegas2024.sched.com https://etlslasvegas2024.sched.com/event/1hAXH/solutions-hub-sonatype-experience-sonatype-sbom-manager View more about this event at Enterprise Technology Leadership Summit Las Vegas 2024
Reminder: The breakout sessions are starting in 5 minutes. Start navigating your way to whichever session you’re attending. https://devopsenterprise.slack.com/files/UATE4LJ94/F07H7N10D62/timer.png
CLEAR talk: “Re-enroll 6.8mm members in 2 months at physical locations” Wow! 🤯
Shout-out to LaunchDarkly in the CLEAR talk for making the launch super smooth.
@austinhpuckett “Training our business partners on our engineering tools.” I’m wondering how receptive those business-oriented users were to learning those tools? That can be a scary thing for some. I think it also depends on the health of the relationship between engineering and business teams… What was that like?
It definitely require a bit of give and take from both parties. But at the end of the day we developed a tight runbook, worked in tandem through the first few configuration changes and then when we took the training wheels off taught them how to page us if anything went wrong.
It comes back to the shared understanding piece though - they knew very early on the value of this tool and were invested in learning. It helps that they were getting phone calls that the lanes were backed up and rather than tap us they could make the change so they also felt empowered
@james047’s assertion in this session - Cloud was to Ops what AI will be to Security Agree? Skeptical?
skeptical, seems to generic to me, I can change that also to: AI is to Ops what Cloud was to Ops 10 years ago. Would that not also be a true statement?
Me answering that from a perspective, where we found out together with AWS that simply certain things run more cost-effective on-premise and are too expensive in the cloud (still using kubernetes though...)
@james047 "penalties of shift left" — So true. What I learned from Dr Steve Spear — he didn't buy the notion that it was possible to shift left. Made analogy of plant safety people: "you can't teach/train every plant floor worker safety — safety teams need to make tools that plant floor workers can use in their daily work." So too with infosec. "shift-left" is actually the ideally decoupling of infosec from dev work — infosec makes tools that devs can use to get fast and frequent feedback, so they can generate better infosec outcomes.
AI does well: these should be good for infosec • transform, summarize, info extract, generation
Of course, you still need to know what questions to ask.
Holy crap. GPT4 beats SAST 5/5 times. Of course — GPT4 even beat MedPaLM, specialized model for healthcare — the power of these larger models beating more smaller (and older), specialized models is amazing.
Co-Star prompting: https://medium.com/@thomasczerny/co-star-framework-for-prompt-structuring-7f9a8c221224
Bloomberg gave an amazing presentation on how they trained a GPT3.5 class model focusing on finance (costing millions of dollars) — then discovers that GPT4 beat that model. Had fascinating conversation with Devlin from Vanguard, who is presenting tomorrow, on this.
There was a similar outcome two years ago at Google. There was a huge effort to build a specialized model for healthcare and medicine. Turned out the general model (Gemini) beat it every time.
Indeed, Paige Bailey from Google will be talking about this tomorrow! 🎉 (Thanks to @amandalewis @nathen.harvey !!)
>> The development and training of BloombergGPT, a large language model with 50 billion parameters, involved significant costs. Bloomberg spent over $10 million on training this AI model, which was specifically targeted at financial applications[4]. The training process required approximately 1.3 million hours of GPU time on Nvidia's A100 GPUs, utilizing Amazon's AWS cloud services[2]. Additionally, the training involved the use of 512 A100 GB GPUs over a period of 49 days[3]. These figures reflect the substantial investment Bloomberg made to create a domain-specific AI model for financial tasks. >> > Citations: > [1] https://www.ankursnewsletter.com/p/bloomberggpt-making-finance-smarter > [2] https://www.hpcwire.com/2023/04/06/bloomberg-uses-1-3-million-hours-of-gpu-time-for-homegrown-large-language-model/ > [3] https://www.linkedin.com/posts/rajistics_what-it-takes-to-build-bloomberggpt-and-activity-7089609531961348096-iIAs > [4] https://www.threads.net/%40ethan_mollick/post/C46AfItO8RS?hl=en > [5] https://www.reddit.com/r/singularity/comments/128v9iz/introducing_bloomberggpt_bloombergs_50billion/ > [6] https://news.ycombinator.com/item?id=39809177 > [7] https://www.bloomberg.com/tosv2.html?url=L2NvbXBhbnkvcHJlc3MvYmxvb21iZXJnZ3B0LTUwLWJpbGxpb24tcGFyYW1ldGVyLWxsbS10dW5lZC1maW5hbmNlLw%3D%3D&uuid=3df5cb00-5069-11ef-89be-a8bf00c2fe45 > [8] https://www.linkedin.com/posts/pascalbiese_bloomberg-paid-1-million-in-training-costs-activity-7110382128080576512-8o6V > > Thank you Perplexity!
From @james047 https://www.dryrun.security/resources/csa-guide
Don’t Miss Cloudflare’s Demo at 12:30pm today at the Solutions Hub located in the Expo Hall. https://etlslasvegas2024.sched.com/event/1hAXl/solutions-hub-cloudflare-what-if-the-10x-developer-was-all-the-developers-you-hired-along-the-way In this live coding demo we will explore what the future can look like when you can chat with your code and generate scaffolding, fix bugs, and even pivot to new use cases. (edited) http://etlslasvegas2024.sched.com https://etlslasvegas2024.sched.com/event/1hAXl/solutions-hub-cloudflare-what-if-the-10x-developer-was-all-the-developers-you-hired-along-the-way View more about this event at Enterprise Technology Leadership Summit Las Vegas 2024
Don’t Miss Sonatype’s Demo at 1:00pm today at the Solutions Hub located in the Expo Hall. https://etlslasvegas2024.sched.com/event/1hAXH/solutions-hub-sonatype-experience-sonatype-sbom-manager Discover how SBOM Manager can streamline and automate the auditing, distribution, and monitoring of SBOMs at scale, ensuring compliance with the latest regulations. http://etlslasvegas2024.sched.com https://etlslasvegas2024.sched.com/event/1hAXH/solutions-hub-sonatype-experience-sonatype-sbom-manager View more about this event at Enterprise Technology Leadership Summit Las Vegas 2024
Reminder: The breakout sessions are starting again in 5 minutes. Start navigating your way to whichever session you’re attending. https://devopsenterprise.slack.com/files/UATE4LJ94/F07H7N10D62/timer.png
Oscar Health: @chawthorne : Our NPS is 66, which is awesome. Most insurance companies have NPS score of... zero. 😆
Interesting trivia fact: earlier in her career, Clare was an auditor! Go audit! 🎉
"We embed ProductOps into the teams; it's most controversial, because it's often viewed as expensive. Especially compared to people who merely coach the teams. But this is actually our superpower."
There you have it — @chawthorne , former auditor, talking about "minimum viable process"
Melissa Perri: Product Operations: How successful companies build better products at scale https://a.co/d/ajBoDhh
"For years, we've been expensing our infrastructure spend.. we started capitalizing those costs" (!!!). 🎉 🎉 🎉
@chawthorne I love that you insisted all run work had to be included in the product roadmap, and it's become a forcing function in run investment 👏
But with Product Ops as a bridge between Product and Engineering, does your team reporting line exist outside both functions? :thinking_face:
Yup - my boss’s other direct reports are our Head of Product & Design, our Head of Product Eng and our head of Infrastructure Eng. Tech Ops is a partner alongside!
Show collaborative intent by always leaving some money on the table!
Statement of objective (descriptive) rather than Statement of Work (prescriptive)
For anyone joining learning sprint: Achieving Clarity in Structure and Strategy 2:40 we are now in Capri-6 If you can bring a hot drink from the break, there’s a extra learning opportunity for you from :flag-au:🦘
Reminder: The plenary sessions are starting again in 5 minutes. Start making your way back to the Azure Ballroom. https://devopsenterprise.slack.com/files/UATE4LJ94/F07H7N10D62/timer.png
⚕️ Please welcome David Wright, Founder & CEO, Disruptive Innovations, who will present: The Fallout from Disjointed Hospital Systems and Siloed Knowledge as a Result of Mergers and Acquisitions
What a day! I can't wait for the last bit of programming today. I saw an early view of @jason.cox's slides and I am so excited to hear the full talk.
2Key stakeholders: chief radioloist, chief medical officer, chief operations officer, manager of radiology operations, CIO
PACS and CPACS: picture and archiving systems, digital image management system. Flickr, but for doctors. 😂
No, that's not right. PACS/CPACS implements all the critical workflows for all the different clinician specialities.
What did we find? "Everything they told us was wrong. We expected 3 PACS systems, we found 5..."
Once upon a time when I was writing and delivering end to end appplication and infrastructure monitoring software I helped monitor the image recall/retrieval process for a health care company so they could show how much doctor and patient time and morale and money was wasted with an old, slow app. This is taking me back.
Problem #1: clinicians (radiologists, etc) can't access images quickly; Problem #2: clinicians in hospital rooms: images take 10m to load — way too long to use during daily rounds.
silo'd teams and silo'd systems leads to complacency with systems and no longer using the systems for them.
There is actually a program I support with a product called “Keep the lights on”
I’m guessing that every project that implemented one element of that mess was a “successful” project though
Lots of healthcare in my family, these issues are rampant across healthcare.
When no one can solve the “whole” problem, they over engineer their part of the problem, making the whole much worse. But feeling locally really good about their firewall config!
What I found so interesting about this case study is to solve this type of problem, you need expertise in both Layer 3 (org wiring: coordination mechanisms across teams) and Layer 2 (deep and wide technical expertise: VPN, firewalls, storage servers).
it's hard to fix things when you don't have visibility into what goes wrong and where. Observability is key to finding and fixing the issues (this is true of culture/organizational alignment and not just technology).
Almost without fail, in regulated ,large enterprises like this and financial/banking, the ability to coordinate all the necessary SME to fix this is never prioritized or taken seriously...until it becomes a costly emergency. Then vendors are called. Leadership issues.
🌺 A warm aloha to Tracy Behler, Senior Director - Online Experience, and Kiyoshi Kusachi, Senior Director - IT, from Hawaiian Airlines, here to present Modernizing Business-Critical Systems Without Negatively Impacting Customers at Hawaiian Airlines
Enterprise Tech Airline Leadership Summit ✈️ ❤️
And we all thank you for flying us to the conference safely, with modern, effective and stable systems 🍻
"Unfortunately, we didn't have app capability to cancel flights; it crushed all call centers. It went on for weeks"
Each island had different travel restrictions, and they were constantly changing.
"Each island had different travel restrictions, which we had to enforce, and they were constantly changing."
"...we felt we had reached cruising altitude, until six months..." "Our company made a decision to replace our PSS, from SABRE to Amadeus — like a heart and lung transplant"
Capability flags FTW 🚩 https://trunkbaseddevelopment.com/feature-flags/
Reservations/checkin, flight ops, travel agencies & OTA, revenue management, loyalty programs, payments, call center...
I flew Hawaiian earlier this year. Directly booked through the app and everything went great. Amazing work transforming such a dynamic operation 🎊
"Our customers were super pissed. Customer satisfaction dropped to lowest I've ever seen"
"To hit our date, continuous delivery was not a nice to have. It was a must have."
Now I want to hear about that replacement decision and what led to them being in this situation at all... :thinking_face:
I hear that sometimes these things are sometimes caused by vendors who put customers are situations where "contract renewal is coming up — we've got to switch now, or get locked into another 10 year contract". Cc @paul.gaffney
The Sabre vs Amadeus market battle, as I understand it, has been epic and I wouldn’t be surprised if something like Gene describes happened
Kudos to them for being able to make the switch at all! I've seen that dynamic where it did not come out with this happy ending. Very impressive. 🙂
We do seem to be hearing that perhaps the only positive feature of ridiculous “deadlines” is that teams have to embrace more rapid delivery techniques…
"Create the blueprint for where you want to be" this is so important. it reminds me of the quote: "If you don't know where you're going, all roads will get you there" - Lewis Carroll
In retrospective, I wonder if the dip on CX could be the requirements of the MVP ? :thinking_face:
And now, let's welcome Jason Cox, Director, Platform & SRE, and Alexi Varanko, VP, Cloud, Infrastructure & SRE from The Walt Disney Company, here to present: Disney Global SRE - Funding the Magic for Embedded Teams
4Always a highlight to see what 🪄 that @jason.cox and team put together for their talk track.
Omg, I could not imagine @jason.cox having a boss in real life!!! I thought it was Walt himself!!!
There you go @topo.pal! Conway’s Law made its appearance…
TPP Graphic Novel galleys available tomorrow night, friends!!!
ITIL v3 was the 26 process framework - 26 queues. I have talked to multiple orgs who tried to actually implement it in the 2012-15 timeframe. I'm not aware of any successes.
Consultants were well paid though, so they considered it a success
https://www.forbes.com/sites/lisabodell/2020/08/28/futurethink-forecasts-t-shaped-teams-are-the-future-of-work/ "They have a working knowledge of other team members' roles and responsibilities, which allows them to collaborate effectively. Collaborative Strengths: T-shaped team members excel in cross-functional collaboration."
https://images.app.goo.gl/vs39SvxW3MM9trxM8 I think this image answers it the best
@biehnam a T shaped teams have depth in one or more areas breadth across many areas. Ideally when we put this team together they have the skills to complete a wide variety of needs.
this should be "community of <insert>" instead. it's a total mindset shift.
this is not a cut n dry subject. Application and enterprise context matter...alot.
I love this "thank you, Jason" reel from people across Disney. 😂😂😂