# Mehmet Deniz Birlikci

![Mehmet Deniz Birlikci](/resources/cognition/portraits/mehmet-deniz-birlikci.webp)

Research @ Cognition | Prev Amazon AGI, Mercedes F1, Carnegie Mellon | Neo Scholar  
Cognition  
San Francisco, California, United States

## At Cognition

- Role: Research (Research)
- Joined: May 2026
- How they joined: Directly
- Before Cognition: Research, Amazon AGI SF Lab, Amazon
- Overlapped before Cognition with other Cognition people at: Amazon (https://www.autumn.ai/resources/cognition/worked-together)

## Identity
- LinkedIn: https://www.linkedin.com/in/mehmet-deniz-birlikci
- X: @denizbirlikci
- GitHub: https://github.com/deniz-birlikci
- Website: https://denizb.io/
- Domain: cognition.ai

## Notes
- On SWE-2 launch he wrote verbatim: "SWE-2 is our strongest model yet and our first with different reasoning efforts.

We’ve show that careful RL design can advance the Pareto of cost and intelligence.

On FrontierCode, we’re within one point of fable while 64% cheaper. https://t.co/MWiujb0y17" (https://x.com/denizbirlikci/status/2098094057826459976, https://t.co/MWiujb0y17)
- On Cognition's fundraise he wrote verbatim: "Cognition’s run-rate revenue has grown from $492M to almost $900M since May -- and we're just getting started.

Today's fundraise allows us to be more ambitious, bring exceptional people on, and scale our bets further.

Come join us!" (https://x.com/denizbirlikci/status/2097374440938254788)
- Listed as a co-author of Cognition's 'Introducing FrontierCode' coding eval (June 2026) alongside Eric Lu, Ben Pan, Sam Lee, Ray Wang, Rohan Choudhury, Fermi Ma, TC Qin, Carlo Baronio and S. Alberti, and of 'FrontierCode 1.1' (July 7, 2026); his core Cognition research team is visible from these bylines. (https://cognition.com/blog/frontier-code)
- Joined Cognition in Research in May 2026 after ~2 years at Amazon AGI SF Lab (Jul 2024-May 2026); the timing puts him there just before the SWE-2 and FrontierCode pushes he now posts about. (https://www.linkedin.com/in/mehmet-deniz-birlikci)
- Current role is Research at Cognition. (https://www.linkedin.com/in/mehmet-deniz-birlikci)
- Listed as an author of Cognition's 'SWE-1.7: Frontier Intelligence at a Fraction of the Cost' (Ben Pan*, Carlo Baronio*, Rohan Choudhury, Eric Lu, Ryan Kim, Deniz Birlikci, TC Qin, ...). (https://cognition.com/blog/swe-1-7)
- Personal site (written while at Amazon): he led reinforcement learning at Amazon AGI Labs on the Nova Act browser agent, helped take it 'from zero to a full AWS service' powering Alexa+ for 1M+ users, and 'built out and led the RL scaling team'; earlier he 'ran multimodal post-training for our research preview'. (https://denizb.io/)
- Current role is Research at Cognition. (https://www.linkedin.com/in/mehmet-deniz-birlikci)
- On SWE-2 demand he wrote verbatim: "SWE-2 has shown us every second counts when you ship a model people love.

We were able to secure additional compute in record time, and now, we're making SWE-2 free in Devin Cloud until Oct 8th.

It's a sweet (but cautionary) tale that emphasizes just how important compute planning is in this age. Compute is scarce and deals often take time. I'm very glad we've been able to move fast and make this happen." (https://x.com/denizbirlikci/status/2101048845845283197)
- On SWE-2 cost efficiency he wrote verbatim: "I think this is our craziest result. SWE-2 is ~80% cheaper than SWE-1.7 model despite being 3x larger.

It shows that a larger model can be more cost-efficient per task, even if it's more expensive per token, because it runs more efficiently.

Astra's cheaper than Sol. Fable 5.1 is cheaper than Opus. Now SWE-2 is cheaper than SWE-1.7." (https://x.com/denizbirlikci/status/2098107227798487274)
- Recorded Episode 4 of Amazon's 'Making a Mind' podcast, 'Improving Agent Reliability with Reinforcement Learning', with host Danielle Perszyk, covering the RL stage for Nova Act (environments, verifiers, RL infrastructure); the episode frames 'Reliability, not accuracy' as the bottleneck for web agents. (https://www.youtube.com/watch?v=FbEISiKZvQc)
- LinkedIn post on the Nova Act launch, verbatim: "What a joy it is to be part of the team that made Nova Act happen! ... It's been a huge research & storytelling milestone for me. I got to both: 1. learn from amazing researchers and train our model 2. design a vision for our launch and bring it to life"; he owned launch storytelling as well as training. (47e0c)
- Joined Adept as part of its incoming new-grad class in July 2024 and moved to Amazon with the Adept acquihire (Amazon hired Adept's founders), which is how he landed at Amazon AGI SF Lab. (https://techcrunch.com/2024/06/28/amazon-hires-founders-away-from-ai-startup-adept/)
- On Decagon using Devin he wrote verbatim: "Optimizing inference is now much easier with agents -- making it a no-brainer for high-TPM orgs.

Decagon used Devin to do so.

When I met @_pavidhiman and @cyrusasg a month ago, it was clear how killer Decagon research team was. Expect to see much more bangers from them soon." (https://x.com/denizbirlikci/status/2115584521223467480)
- On Devin Fusion he wrote verbatim: "Devin Fusion now uses SWE-2 to give you the same performance as Fable 5.1 Max at 36% less cost.

There's a lot of juice to squeeze both on research and the harness side. Cognition is in a unique position being able to do both." (https://x.com/denizbirlikci/status/2098454179249594559)
- Cited as an author of Cognition's SWE-1.7 in the SWE-2 launch post's references (B. Pan, C. Baronio, R. Choudhury, E. Lu, R. Kim, D. Birlikci, ...). (https://cognition.com/blog/swe-2)
- On coding-model UX he wrote verbatim: "I'm convinced scope violations are the biggest UX problem in coding models right now.

It's remarkable how much frontier models fail to write simple implementations, and crazier how much headache it causes.

We were early w/ FrontierCode; I expect more evals to capture this soon" (https://x.com/denizbirlikci/status/2093501462588670119)
- Public headline says "Research @ Cognition | Prev Amazon AGI, Mercedes F1, Carnegie Mellon | Neo Scholar". (https://www.linkedin.com/in/mehmet-deniz-birlikci)
- Interviewed at NeurIPS 2025 as Member of Technical Staff at Amazon AGI Labs on the cutting edge of web agents ('Why 80% Accuracy Fails in AI Agents'), a second public talk on agent reliability. (https://www.youtube.com/watch?v=6Aq1bZ3vJuY)
- First paper: 'GeSTICS: A multimodal corpus for studying gesture synthesis in two-party interactions with contextualized speech' (IVA 2024, ACM), co-authored with Gaoussou Youssouf Kebe, Auriane Boudin, Ryo Ishii, Jeffrey M. Girard and Louis-Philippe Morency. (https://affcom.ku.edu/people/staff/girard_cv.pdf)
- In summer 2023 he joined Louis-Philippe Morency's MultiComp Lab at CMU's Language Technologies Institute as a research assistant, working with LP and Youssouf Kebe on co-speech gesture generation. (07477)
- Public headline says "Research @ Cognition | Prev Amazon AGI, Mercedes F1, Carnegie Mellon | Neo Scholar". (https://www.linkedin.com/in/mehmet-deniz-birlikci)
- He wrote verbatim: "Seeing @AryaGrayeli, @RonNachum, and @pranavravella grow this company to from an idea to a critical intelligence layer for companies has been amazing.

As intelligence increases, companies will need AI systems that deeply understand them." (https://x.com/denizbirlikci/status/2097371400235634978)
- Reposted Matt Bergland's recap that Cognition in one week 'acquired @anhangz @yunpark93 / TierZero' and released Devin Outposts, and Walden Yan's note that Cognition has only ~70 engineers. (https://x.com/denizbirlikci/status/2080397397138763860)
- On Devin Outposts-era launch he wrote verbatim: "We love Devin and want you to be able to run it inside any execution environment." (https://x.com/denizbirlikci/status/2079619600417952255)
- Deniz recently graduated from Carnegie Mellon with a Bachelor's in AI and is a Neo Scholar. He enjoys merging research with building innovative products, something he realized while co-founding a startup developing a multimodal search engine for fashion e-commerce. Previously, he conducted research with CMU's Multimodal Lab, deployed AutoML pipelines at Mercedes F1 team, and taught over 1,000 students as a head TA; this comes from the profile summary and preserves the person's own positioning. (https://www.linkedin.com/in/mehmet-deniz-birlikci)
- Interviewed by NBC News (article by Victoria Feng) about CMU's BS in AI program; he said BSAI opened doors 'all the way from interning at Mercedes F1 team to now joining Amazon AGI'. (4cb44)
- ObviousAI was a 2024 Microsoft Imagine Cup semifinalist; Neo founder Ali Partovi commented 'Small world! I'm looking forward to judging the finals next month.' on his announcement. (https://www.linkedin.com/posts/mehmet-deniz-birlikci_announcing-the-2024-imagine-cup-semifinalists-activity-7186579998630174723-L2OH)
- One of two CMU students awarded the Swartz Center for Entrepreneurship startup stipend (2023), and one of 13 Swartz Center Innovation Scholars, which funded the ObviousAI work. (07477)
- Interned on Race Strategy at the Mercedes-AMG F1 team, where per his profile he deployed AutoML pipelines. (https://denizb.io/)
- Deniz recently graduated from Carnegie Mellon with a Bachelor's in AI and is a Neo Scholar. He enjoys merging research with building innovative products, something he realized while co-founding a startup developing a multimodal search engine for fashion e-commerce. Previously, he conducted research with CMU's Multimodal Lab, deployed AutoML pipelines at Mercedes F1 team, and taught over 1,000 students as a head TA; this comes from the profile summary and preserves the person's own positioning. (https://www.linkedin.com/in/mehmet-deniz-birlikci)
- Previous roles include Research at Cognition; Neo Scholar at Neo; Research, Amazon AGI SF Lab at Amazon; Member of Technical Staff at Adept; University Founder Fellow at Unusual Ventures; Co-Founder at ObviousAI; Multimodal AI Researcher at Carnegie Mellon University - School of Computer Science - Language Technologies Institute; Head Teaching Assistant for 15-112 (Fundamentals of Programming) at Carnegie Mellon University. (https://www.linkedin.com/in/mehmet-deniz-birlikci)
- Unusual Ventures University Founder Fellow (Feb-Jul 2024) and Neo Scholar (from Oct 2023), two early-career founder networks. (https://www.linkedin.com/in/mehmet-deniz-birlikci)
- Head Teaching Assistant for CMU 15-112 (Fundamentals of Programming), teaching 1,000+ students. (https://denizb.io/)
- Previous roles include Research at Cognition; Neo Scholar at Neo; Research, Amazon AGI SF Lab at Amazon; Member of Technical Staff at Adept; University Founder Fellow at Unusual Ventures; Co-Founder at ObviousAI; Multimodal AI Researcher at Carnegie Mellon University - School of Computer Science - Language Technologies Institute; Head Teaching Assistant for 15-112 (Fundamentals of Programming) at Carnegie Mellon University. (https://www.linkedin.com/in/mehmet-deniz-birlikci)
- Congratulated the OpenHands (All Hands AI) team on their 32B coding-agent model in public comments (Mar 2025). (https://www.linkedin.com/posts/all-hands-ai_today-were-excited-to-make-two-big-announcements-activity-7312528074305150977-ghcm)
- Runs a weekly AI reading group in SF among friends (tweet Mar 2025), a recurring community touchpoint. (https://x.com/denizbirlikci?lang=en)
- Location is San Francisco, California, United States. (https://www.linkedin.com/in/mehmet-deniz-birlikci)
- Location is San Francisco, California, United States. (https://www.linkedin.com/in/mehmet-deniz-birlikci)

## Experience
- Research at Cognition (May 2026 - Present)
- Neo Scholar at Neo (Oct 2023 - Present)
- Research, Amazon AGI SF Lab at Amazon (Jul 2024 - May 2026): zero to one for Nova Act, Amazon's browser use agent
led the RL scaling team
- Member of Technical Staff at Adept (Jul 2024 - Jul 2024): Part of Adept’s incoming new grad class. Moved to Amazon with the acquihire.
- University Founder Fellow at Unusual Ventures (Feb 2024 - Jul 2024)
- Co-Founder at ObviousAI (Jul 2023 - May 2024): Building multimodal AI search for fashion e-commerce.
- Multimodal AI Researcher at Carnegie Mellon University - School of Computer Science - Language Technologies Institute (May 2023 - May 2024): Researching interactions between gesture and speech. Published my first paper into ACM.
- Head Teaching Assistant for 15-112 (Fundamentals of Programming) at Carnegie Mellon University (Jan 2021 - May 2023): Shared my passion of programming with 1000+ other students. Lead the creation of the experimental version of the course (git branch) and made it the norm in three semesters (git merge).
- Software Engineer Intern at Mercedes-AMG Petronas Formula One Team (Jun 2022 - Aug 2022): Architected and deployed an AutoML pipeline, helping all engineers train/deploy AI models over 70,000 data channels with a single API call. Race strategy team.
- Co-Founder at Ludens Store (Oct 2016 - Aug 2017): First entrepreneurship experience. Wore multiple hats: building the website payment infrastructure, handling delivery logistics, and optimizing our digital advertising. 

Ludens Store was an e-commerce company selling sweet-filled gift boxes. We supported same-day delivery and a suite of personalization options. At it's peak, we were doing 5 transactions a day.

## Education
- Carnegie Mellon University - Bachelor of Science - BS, Artificial Intelligence: Head Teaching Assistant for 15-112, Swartz Center for Entrepreneurship Innovation Scholar, Sigma Phi Epsilon

Swartz Center for Entrepreneurship Innovation Scholar:
One of 13 students chosen for a 2-year entrepreneurial program. Includes a proseminar in entrepreneurial leadership, a Silicon Valley Trek with grants, a stipend for own startups, and a robust mentorship program.

## Sources
- https://www.linkedin.com/in/mehmet-deniz-birlikci
- https://denizb.io/
- https://cognition.com/blog/frontier-code
- https://cognition.com/blog/swe-1-7
- https://x.com/denizbirlikci?lang=en
- https://www.linkedin.com/posts/all-hands-ai_today-were-excited-to-make-two-big-announcements-activity-7312528074305150977-ghcm
- https://affcom.ku.edu/people/staff/girard_cv.pdf
- https://www.linkedin.com/posts/danielleoscillations_a-system-that-succeeds-once-is-a-demo-a-activity-7424874158834147328-DYAQ
- https://www.youtube.com/watch?v=6Aq1bZ3vJuY
- https://cognition.com/blog/swe-2
- https://www.youtube.com/watch?v=FbEISiKZvQc
- https://x.com/moritz_stephan/status/1932498121910780241
- https://www.linkedin.com/posts/mehmet-deniz-birlikci_announcing-the-2024-imagine-cup-semifinalists-activity-7186579998630174723-L2OH
- https://x.com/denizbirlikci
- https://github.com/deniz-birlikci
- https://www.linkedin.com/in/mehmet-deniz-birlikci?miniProfileUrn=urn%3Ali%3Afsd_profile%3AACoAACFRQ3QBDlw4xSu_YdcxlxIxMSAf5ExWj-Y
- https://www.linkedin.com/posts/mehmet-deniz-birlikci_earlier-this-year-i-had-two-goals-for-how-activity-7068075627538477056-QN-k
- https://puck.news/ai/2025/8/amazons-ai-lab-is-still-chasing-agi
- https://www.amazon.science/blog/new-making-a-mind-podcast-explores-science-of-intelligence
- https://art19.com/shows/making-a-mind/episodes/255393a6-87a3-4375-b11e-9afa5ce8e39c
- https://orcid.org/0009-0005-0809-3656
- https://x.com/denizbirlikci/status/2098094057826459976
- https://t.co/MWiujb0y17
- https://x.com/denizbirlikci/status/2097374440938254788
- https://x.com/denizbirlikci/status/2101048845845283197
- https://x.com/denizbirlikci/status/2098107227798487274
- 47e0c
- https://techcrunch.com/2024/06/28/amazon-hires-founders-away-from-ai-startup-adept/
- https://x.com/denizbirlikci/status/2115584521223467480
- https://x.com/denizbirlikci/status/2098454179249594559
- https://x.com/denizbirlikci/status/2093501462588670119
- 07477
- https://x.com/denizbirlikci/status/2097371400235634978
- https://x.com/denizbirlikci/status/2080397397138763860
- https://x.com/denizbirlikci/status/2079619600417952255
- 4cb44


Researched with Autumn (https://www.autumn.ai) from public sources. Updated 2026-11-04. Profile: https://www.autumn.ai/resources/cognition/people/mehmet-deniz-birlikci · Everyone at Cognition: https://www.autumn.ai/resources/cognition/people