Sign In

Alex Abramov

Engineering Leader | Metropolis | ex-Meta | Observability, Cloud Cost Management

Alex Abramov is the Vice President of Engineering at Metropolis Technologies, a company focused on leveraging artificial intelligence and computer vision to modernize parking solutions. He has a rich background in engineering leadership, having previously held significant roles such as Senior Engineering Org Leader at Meta (formerly Facebook) and Senior Director of Engineering at Jet. His experience also includes positions as VP of Engineering at CADRE and Bottomline Technologies, among others.12

Abramov holds a Bachelor of Science degree in Computer Science from Brown University, which has provided him with a solid foundation for his career in technology.1 His expertise spans areas such as AI, machine learning infrastructure, and cloud cost management, where he has been instrumental in driving innovation and leading high-performing teams.2

In his current role at Metropolis Technologies, Abramov is responsible for spearheading engineering initiatives that enhance the company's technological capabilities and operational efficiency. His leadership is characterized by a focus on team development and problem-solving on a global scale.1

Highlights

May 15 · linkedin

𝐆𝐏𝐔 𝐏𝐫𝐨𝐟𝐢𝐥𝐢𝐧𝐠 𝐂𝐮𝐬𝐭𝐨𝐦𝐞𝐫 𝐒𝐭𝐮𝐝𝐲 𝐢𝐧 𝐆𝐞𝐧 𝐀𝐈 𝐬𝐩𝐚𝐜𝐞

Good morning LinkedIn community! It’s been about 2 months since my last post (https://lnkd.in/e5-4ibWH ) and I’d like to provide an update on what I’ve been up to. As I mentioned in my prior post, I have been looking deeper at how GPU Profiling work, inspired by this research (https://lnkd.in/e8sScCMN) I worked on with the Strobelight team at Meta, can be applied to the industry at large.

We’ve conducted market research by reaching out to ~200 companies out of which we’ve talked to approximately two dozen. The results are linked below.

𝐖𝐡𝐲 𝐚𝐫𝐞 𝐰𝐞 𝐬𝐡𝐚𝐫𝐢𝐧𝐠 𝐨𝐮𝐫 𝐫𝐞𝐬𝐞𝐚𝐫𝐜𝐡 𝐡𝐞𝐫𝐞? Many early-stage startups choose to stay in a stealth mode during the early stage research. However, we believe this is not the right approach in this case. The AI/ML space is developing rapidly and the industry players are telling us that the value of time is high and rapid progress is a must.

We believe that the progress in the Gen AI industry is significantly slowed down by the issues companies discover with AI Infra. Our goal in sharing our market research is to raise awareness and see if there is an interest in the community for closer collaboration. Given the findings referenced below, we truly believe that leveraging novel techniques in GPU Profiling is the path to accelerating the industry forward. Please, DM me if you want to collaborate or exchange ideas!

𝐓𝐋;𝐃𝐑 𝐨𝐧 𝐭𝐡𝐞 𝐟𝐢𝐧𝐝𝐢𝐧𝐠𝐬 1-We are in an AI arms race and the industry players feel the short-term value of time is extremely high. 2-Companies running large distributed ML Training workloads on clusters of GPUs lose weeks of training time per model and incur $mils in extra costs due to various issues related to GPU cluster configuration or the nature of training workload. 3-Model Builders are the ones experiencing the brunt of the infra pain, not the Infra Providers. Furthermore, the way that economic models are currently aligned means Infra providers do not have meaningful incentives to solve this problem for model builders.

See the link to the study below.

𝐖𝐡𝐚𝐭’𝐬 𝐍𝐞𝐱𝐭 I’ve teamed up with Charlie Kroll to interview key players in the Gen AI/MLOps market to get a signal on the commercial value of our ideas. The thought was if we’d find that the signal is strong, we’d try to secure a customer pilot.

As a result of our market research, we have a pipeline of potential pilots lined up. 𝐍𝐞𝐱𝐭, 𝐈’𝐦 𝐥𝐨𝐨𝐤𝐢𝐧𝐠 𝐭𝐨 𝐛𝐫𝐢𝐧𝐠 𝐚 𝐭𝐞𝐜𝐡𝐧𝐢𝐜𝐚𝐥 𝐜𝐨-𝐟𝐨𝐮𝐧𝐝𝐞𝐫 𝐰𝐢𝐭𝐡 𝐝𝐞𝐞𝐩 𝐞𝐁𝐏𝐅 𝐭𝐞𝐜𝐡 𝐞𝐱𝐩𝐞𝐫𝐢𝐞𝐧𝐜𝐞 𝐭𝐨 𝐤𝐢𝐜𝐤-𝐬𝐭𝐚𝐫𝐭 𝐭𝐡𝐞 𝐩𝐢𝐥𝐨𝐭(𝐬) 𝐰𝐢𝐭𝐡 𝐦𝐞. If you are interested in exploring this opportunity, please DM me! I will also be reaching out to my network for referrals. If I’m able to bring a co-founder, then we’ll form a company and will quickly secure a pre-seed funding round.

Mar 19 · linkedin

After almost 5y at Meta, the time has come for me to move on. I’m writing this post to share the news broadly, reflect on the journey at Meta and thank many amazing people I’m going to miss.   some history I joined in 2019 and landed to support a couple of teams in the Observability team of Core Systems group. On one of those teams, we’ve developed great tools such as Tracery, Comparison View, and Landline to empower engineers internally to get to the bottom of tricky perf/reliability/efficiency issues (https://lnkd.in/e8ph--Un). Shoutout to Douglas, Helga, Henry, and others who made this work exciting and fun.

A bit later I helped rebuild our Profilers team. We’ve started with CPU profiling by expanding our customer base from experts to non-experts and enabling $$$ of efficiency wins. And then we’ve pioneered some novel GPU Profiling methods using eBPF tech (https://lnkd.in/e8sScCMN). Shoutout to Jacie, Jonathan, Jordan, Nathan, Riham, and Vlad for being awesome!

In parallel with all of the above, we built our Capacity Tooling team SECT. We’ve mapped out Capacity Management workflows and re-engineered related business processes to make capacity management easy at Meta. As part of this journey, we’ve built Service Console, Capacity Portal, RF, VRP, and other tools that are currently used by 1000s of people internally. Shoutout to Bill, Heike, Matt, Mukund, Peter, Rachelle, and other folks on SECT.

Config Products was a great team I helped bring from IG to Core Systems. It was a lot of fun as the challenges in the area reminded me of config management challenges at Andera. Shout out to Anshul, Joshua, and others.   Last, but not least, I took over the Resource Accounting and Budgeting teams early last year. 2023 was a tough year in many ways. Our teams heroically delivered 90+% fleet coverage attribution to enable resource metering and billing. Shoutout to Akanksha, Jacie, Igor, Richard, Shuyang, and others for all your hard work under pressure.

A word of thanks to the leadership: Jeromy, Yun, Avnish, and Blake. I’ve learned a lot from you.

Special thanks to my XFN partners, who made working with you a joy: Alex, Ariel, Babak, Chandra, Lenni, Valentin, James, Ayichew, Andrew, Haiying, Mark, Andrii, Parth, Patrick and many others (hitting mentions limit :) ).

As I think about my experiences at Meta, I feel gratitude for all the experiences and the people I got to work with. It's a small world, I hope many of us will cross paths again!

what’s next During the next few months, I’ll spend time reconnecting with my network, learning, and brainstorming. I'll be looking at everything from starting a new company to joining an established firm. Specific trends I'm interested in exploring are around general Observability and specific focus areas such as GPU Profiling.

As I'm planning to talk to a ton of people, don't be surprised if you hear from me. Also, please don't hesitate to reach out if you'd like to catch up or throw ideas around.

Related Questions

What are Alex Abramov's key achievements at Metropolis Technologies?
How did Alex Abramov's role at Meta influence his career?
What specific projects has Alex Abramov led at Jet?
How does Alex Abramov's educational background impact his professional approach?
What are the main challenges Alex Abramov has faced in his engineering leadership roles?
Alex Abramov
Alex Abramov, photo 1
Alex Abramov, photo 2
Add to my network

Location

New York City Metropolitan Area