How OpenAI’s engineering org is reshaping teams, roles and workflows w/ Sulman Choudhry #250

59 snips

Mar 10, 2026

Sulman Choudhry, Head of ChatGPT engineering at OpenAI and former product lead at Meta and UberEATS, discusses how engineering teams evolve around missions. He covers fluid mission-driven teams, vertical vs horizontal structures, and the Directly Responsible Individual model. Conversations dive into coordinating many experiments, managers as systems designers, mentoring across levels, and scaling decision-making without bottlenecks.

Ask episode

AI Snips

Chapters

Transcript

Episode notes

INSIGHT

AI Is Collapsing The Inner Loop

AI is collapsing the engineering inner loop by making code generation nearly as cheap as token generation.
Sulman frames inner vs outer loops: inner (code/write/test) is tightening; outer (signals/learning) must catch up to avoid chaos.

ADVICE

Build End To End Teams With Researchers Included

Form end-to-end product teams that include researchers, engineers, designers, PMs and data scientists to pursue both model and product breakthroughs.
Sulman cites ChatGPT Search: a small cross-functional group given weeks to iterate produced the feature.

INSIGHT

Be Explicit About Horizontal Versus Vertical Teams

Maintain explicit horizontal vs vertical teams: product verticals focus on features, horizontals build reusable infra like RTC/voice.
OpenAI created a horizontal RTC team after Advanced Voice revealed lacking expertise.

Get the Snipd Podcast app to discover more snips from this episode

Get the app

In this episode, recorded live at the OpenAI studio, Sulman Choudhry (Head of ChatGPT @ OpenAI) pulls back the curtain on how they structure engineering teams! We talk about shifting from silos to fluid mission-driven teams, vertical vs. horizontal teams, maximizing cross-functional collaboration between research, engineering, product and design. Plus we cover “directly responsible individuals” for high accountability, managers as systems designers, scaling decision-making to prevent leadership from becoming bottlenecks, frameworks for mentoring junior engineers, why “problem framing” is the most critical skill, and how managers can stay close to problems and maintain technical intuition.

ABOUT SULMAN CHOUDHRY

Sulman leads ChatGPT Engineering at OpenAI, driving the development and scaling of one of the world’s most impactful AI products. He pushes the boundaries of innovation by turning cutting‑edge research into practical, accessible tools that transform how people interact with technology. Previously at Meta, Sulman founded and scaled Instagram Reels, IGTV, and Instagram Labs, and helped lead the early development of Instagram Stories.

He also brought MetaAI to Instagram and Messenger, integrating generative AI into experiences used by billions. Earlier in his career, Sulman was on the founding team that built and launched UberEATS from the ground up, helping turn it into a global food delivery platform. With a track record of marrying technical vision, product strategy, and large‑scale execution, Sulman focuses on building products that meaningfully change how people live, work, and connect.

This episode is brought to you by xMatters!

xMatters automates the entire incident lifecycle with their purpose-built AI powered workflow, giving your team the context they need to stop disruptions before they start and minimize resolution times.

Head over to xmatters.com to learn more!

SHOW NOTES:

The Shift to AI-Native Engineering: How AI is collapsing the "Inner Loop" and reshaping engineering team composition (2:48)
Mission-Driven Teams: Moving from traditional functional silos to integrated, problem-centric units (4:45)
Vertical vs. Horizontal Team Architecture: How OpenAI structures specialized horizontal teams (ex. Infrastructure, RTC/Voice) with product verticals (7:04)
Fluid org charts & blurring functional roles: AI-Native teams require proactive mission alignment and coordination over rigid structure (8:48)
The Lifecycle of Problem-Oriented Teams: What happens when a "strike team" solves the problem (10:02)
Maximizing cross functional collaboration between engineering, research, product and design (11:52)
The DRI Framework: Implementing the "Directly Responsible Individual" model for high-velocity accountability (13:32)
Thriving in the "Chaos Factory": Addressing bottlenecks in highly dynamic, high-volume environments (16:02)
Prioritization & "Letting 1,000 Flowers Bloom": How OpenAI decides which AI bets to double down on (19:13)
Scaling Decision-Making: Preventing leadership from becoming the bottleneck as volume increases (21:19)
Knowing when to call it quits on a bet and reallocate talent for maximum impact (23:29)
The Manager as "Systems Designer": Shifting the EM role from people logistics to technical orchestration (24:49)
The Barbell Talent Strategy: Optimizing for innovation by pairing "super seniors" with "super juniors" (28:10)
Mentorship in the AI Age: How to coaching junior engineers when the "cost of code" is approaching zero (30:19)
Technical Intuition for Leaders: Sulman’s frameworks for staying "close to the metal" as a manager (33:17)
Cultivating Judgment: Why "Problem Framing" is the most critical skill for the modern engineer (37:01)
Rapid fire questions (38:59)

LINKS AND RESOURCES:

99% Invisible](https://99percentinvisible.org/): The design and architecture podcast Sulman has followed for over a decade.
The Invisible Cow Tunnels of Chicago](https://99percentinvisible.org/episode/cow-tunnels/): A specific episode of 99% Invisible mentioned by Patrick.

This episode wouldn’t have been possible without the help of our incredible production team:

Patrick Gallagher - Producer & Co-Host

Jerry Li - Co-Host

Noah Olberding - Associate Producer, Audio & Video Editor https://www.linkedin.com/in/noah-olberding/

Dan Overheim - Audio Engineer, Dan’s also an avid 3D printer - https://www.bnd3d.com/

Ellie Coggins Angus - Copywriter, Check out her other work at https://elliecoggins.com/about/

Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.