# Joe Li

![Joe Li](/resources/cognition/portraits/joe-li.webp)

Stanford EE student; RL for coding models (ex-Cognition SWE-1.6 era intern, ex-Nous Research NousCoder-14B lead); USACO Finalist  
Stanford University  
Cupertino, California, United States

## At Cognition

- Role: Former ML Research Intern (Research)
- Joined: Jan 2026
- How they joined: Directly
- Before Cognition: ML Research, Nous Research
- Now: Stanford University (https://www.linkedin.com/in/joe-li-08791120b)
- Overlapped before Cognition with other Cognition people at: Stanford University (https://www.autumn.ai/resources/cognition/worked-together)

## Identity
- LinkedIn: https://www.linkedin.com/in/joe-li-08791120b
- X: @JoeLi5050
- GitHub: https://github.com/JoeLi12345

## Notes
- Nous Research announced NousCoder-14B was 'post-trained on Qwen3-14B by researcher in residence Joe Li using 48 B200s over the course of 4 days', using Atropos and Modal's autoscaler, reaching 67.87% Pass@1 (+7.08% over Qwen's baseline). (https://www.linkedin.com/posts/nousresearch_introducing-nouscoder-14b-a-competitive-activity-7417263170446295040-zbWf)
- ML Research Intern at Cognition in San Francisco from Jan to Mar 2026, described on LinkedIn as 'Training coding models. RL algorithms, infrastructure, data.' (https://www.linkedin.com/in/joe-li-08791120b)
- Cognition's SWE-1.6 Preview (2026-03-01) and SWE-1.6 GA (2026-04-07) were the coding-model RL runs during and right after his internship window; SWE-1.6 is post-trained on the SWE-1.5 base, runs at up to 950 tok/s, and shipped 'generally available in Windsurf'. (https://cognition.com/blog/swe-1-6)
- On 2026-04-07 he retweeted @cognition: 'We're releasing SWE-1.6, our best model in both intelligence & model UX. SWE-1.6 matches our Preview model on SWE-Bench Pro...' His X bio now reads '@Stanford | prev: @Cognition @NousResearch'. (https://x.com/JoeLi5050/status/2041663536691343403)
- At Nous Research (Nov 2024 - Dec 2025) he did pre- and post-training and lists releases nGPT, Hermes 4, and NousCoder-14B. (https://www.linkedin.com/in/joe-li-08791120b)
- Co-author of the Hermes 4 Technical Report (arXiv:2508.18255) with Ryan Teknium, Dakota Mahan, Jeffrey Quesnelle, Chen Guang, Shannon Sands, Karan Malhotra and others. (https://arxiv.org/abs/2508.18255)
- Cognition signed a definitive agreement to acquire Windsurf (IP, product, trademark and brand) on 2025-07-14; he joined Cognition in Jan 2026, after the deal. (https://cognition.com/blog/windsurf)
- 2021-2022 USACO Finalist (Harker, class of 2024), invited to the 2022 USACO summer training camp alongside Harker schoolmate Robert (Bobby) Costin. (https://usaco.org/index.php?page=finalists22)
- Open source: his GitHub JoeLi12345 (bio 'USACO, ML') hosts nGPT, 'an open source reproduction of NVIDIA's nGPT (Normalized Transformer with Representation Learning on the Hypersphere)', with about 112 stars. (https://github.com/JoeLi12345)
- He posted on X about NousCoder-14B: 'Enjoyed working on this a lot! Check out the blog as well.' (https://x.com/JoeLi5050/status/2008642521568866720)
- LLaDA-R1 post-trained LLaDA-8B-Instruct with SFT and RL plus a remasking optimization to adapt diffusion steps to problem difficulty; code at github.com/radi-cho/reasoning-diffusion. (https://www.linkedin.com/posts/joe-li-08791120b_we-took-1st-place-and-won-40k-at-the-mercor-activity-7302221298229751808-6tHZ)
- Co-wrote USACO problems: credited with Larry Xing for December 2022 Gold 'Mountains' and the official analysis (with Benjamin Qi); listed among December 2022 contest helpers with Benjamin Qi, Freddie Tang, Nathan Wang and others. (https://usaco.org/index.php?page=viewproblem2&cpid=1258)
- Public headline says "Stanford | USA IOI Training Camp". (https://www.linkedin.com/in/joe-li-08791120b)
- Researcher at the Hawaii Digital Health Lab (UH Manoa) 2022-2024; first-authored 'A Comparison of Personalized and Generalized Approaches to Emotion Recognition Using Consumer Wearable Devices' in JMIR AI with Peter Washington. (https://ai.jmir.org/2024/1/e52171)
- Princeton Power Electronics Lab (Andlinger Center, Laboratory Learning Program) Jun 2023 - Feb 2024: first-authored a neural-network magnetic core loss paper at IEEE APEC 2024. (https://www.linkedin.com/in/joe-li-08791120b)
- Search-based Planning Lab at Carnegie Mellon University 2021-2022, his first research role, obtained by cold-emailing professors. (https://www.linkedin.com/in/joe-li-08791120b)
- Previous roles include ML Research Intern at Cognition; ML Research at Nous Research; Researcher at Hawaii Digital Health Lab at University of Hawaii at Manoa; Researcher at Princeton Power Electronics Lab at Andlinger Center for Energy and the Environment; Researcher at Search-based Planning Lab at Carnegie Mellon University. (https://www.linkedin.com/in/joe-li-08791120b)
- Location is Cupertino, California, United States. (https://www.linkedin.com/in/joe-li-08791120b)

## Experience
- ML Research Intern at Cognition - San Francisco, CA (Jan 2026 - Mar 2026): Training coding models. RL algorithms, infrastructure, data.
- ML Research at Nous Research (Nov 2024 - Dec 2025): Pre-training and post-training. Released:
- nGPT: Normalized Transformer with Representation Learning on the Hypersphere.
- Hermes 4, a family of hybrid reasoning models.
- NousCoder-14B, a SOTA olympiad coding model of its size post-trained with RL & verifiable execution rewards.
- Researcher at Hawaii Digital Health Lab at University of Hawaii at Manoa (Aug 2022 - May 2024): Biosignal-based emotion recognition using artificial intelligence.
 
First-authored "A Comparison of Personalized and Generalized Approaches to Emotion Recognition Using Consumer Wearable Devices: Machine Learning Study" in JMIR AI.
- Researcher at Princeton Power Electronics Lab at Andlinger Center for Energy and the Environment - Princeton, New Jersey, United States (Jun 2023 - Feb 2024): Princeton Laboratory Learning Program (LLP). Power magnetics modeling.

- First-authored "Investigating the Mutual Impact of Waveform, Temperature, and DC-Bias on Magnetic Core Loss Using Neural Network Models" in IEEE Applied Power Electronics Conference 2024.
- Co-authored "Multi-Material Power Magnetics Modeling with a Modular and Scalable Machine Learning Framework" in IEEE Applied Power Electronics Conference 2024.
- Co-authored "A Simplified Dc-Bias Injection Method for Characterizing Power Magnetics using a Voltage Mirror Transformer" in IEEE Transactions on Power Electronics.
- Researcher at Search-based Planning Lab at Carnegie Mellon University (Jun 2021 - Aug 2022): Computer vision and autonomous planning

## Education
- Stanford University - Bachelor of Science - BS, Electrical Engineering - (Jan 2024 - Jan 2028): Coursework:
Math:
61CM (Real Analysis + Linear Algebra)
122 (Modules and Group Representation Theory)
144 (Topology & Geometry)
147 (Differential Topology)
158 (Probability & Stochastic Differential Equations)
215A* (Algebraic Topology)

Physics:
111 (Partial Differential Equations)
212* (Statistical Mechanics)
330* (Quantum Field Theory)

CS:
109 (Probability)
111 (Operating Systems)
149 (Parallel Computing)
224N* (Natural Language Processing)
336* (Language Modeling from Scratch)

EE:
101A (Circuits I)
102A (Signals and Systems I)
102B (Signals and Systems II)
108 (Digital System Design)
142 (Electromagnetism)
222* (Applied Quantum Mechanics I)
223* (Applied Quantum Mechanics II)
224* (Quantum Control & Engineering)
263* (Matrix Methods: Singular Value Decomposition)
276* (Information Theory)

* indicates graduate coursework.

## Sources
- https://www.linkedin.com/in/joe-li-08791120b
- https://www.linkedin.com/posts/nousresearch_introducing-nouscoder-14b-a-competitive-activity-7417263170446295040-zbWf
- https://usaco.org/index.php?page=finalists22
- https://cognition.com/blog/swe-1-6
- https://cognition.com/blog/swe-1-6-preview
- https://x.com/JoeLi5050
- https://cognition.com/blog/windsurf
- https://x.com/cognition/status/1896738757782696009
- https://x.com/JoeLi5050/status/1896390535033889164
- https://www.linkedin.com/posts/joe-li-08791120b_we-took-1st-place-and-won-40k-at-the-mercor-activity-7302221298229751808-6tHZ
- https://sub.thursdai.news/p/thursdai-jan-8-nvidias-vera-rubin
- https://arxiv.org/abs/2508.18255
- https://github.com/JoeLi12345
- https://usaco.org/index.php?page=viewproblem2&cpid=1258
- https://usaco.org/index.php?page=dec22results
- https://usaco.org/current/data/sol_prob2_gold_dec22.html
- https://www.youtube.com/watch?v=FkQdHdKWAok
- https://archive.teamscode.org/about
- https://ai.jmir.org/2024/1/e52171
- https://x.com/JoeLi5050/status/2041663536691343403
- https://x.com/JoeLi5050/status/2008642521568866720
- https://x.com/JoeLi5050/status/2008953851542663549


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