Jiayi Weng
Born 1998 · Age 27
Research engineer working on reinforcement learning infrastructure and RLHF pipelines; creator/main contributor of Tianshou and EnvPool; educated at Tsinghua University (BSc) and Carnegie Mellon University (MS); research & engineering roles at TSAIL, Sea AI Lab, and OpenAI; multiple publications in RL systems and libraries.
Compare Your Trajectory
See how your career milestones stack up against Jiayi Weng and other industry leaders.
Life & Career Timeline
Started undergraduate studies at Tsinghua University
Began undergraduate (undergrad student) studies at Tsinghua University (expected field: computer science / related).
URBER paper published (TCAD)
Published 'URBER: Ultrafast Rule-Based Escape Routing Method for Large-Scale Sample Delivery Biochips' in IEEE Transactions on Computer-Aided Design (TCAD).
VizDoom AI Competition — 1st place (Single Player Track)
With TSAIL team, proposed environment-aware hierarchical RL architecture and achieved first place in VizDoom AI Competition 2018 Single Player Track.
Personal/side interests: photography, graphics, web security
Engaged as an amateur in photography, computer graphics, and web security (listed on CV).
Built environment-aware hierarchical RL architecture
Research work on environment-aware hierarchical reinforcement learning that led to strong performance in FPS/VizDoom tasks.
Research publication record begins
First research publications (including IEEE TCAD work and early RL work) appearing in 2018–2019 marking start of published research record.
Joined TSAIL (Tsinghua) as research member
Joined the Tsinghua School/Team for AI (TSAIL), working with Prof. Hang Su and Jun Zhu on reinforcement learning research (environment-aware hierarchical RL).
Tianshou: began / became main contributor (PyTorch RL library)
Started contributing significantly to Tianshou, an elegant, flexible, and high-performance PyTorch deep RL library (became main contributor).
Co-authored research on hierarchical RL for FPS games
Paper 'Playing FPS Game with Environment-aware Hierarchical Reinforcement Learning' (IJCAI 2019) — oral presentation.
IJCAI 2019 Oral Presentation — Playing FPS with hierarchical RL
Co-authored and gave an oral presentation 'Playing FPS Game with Environment-aware Hierarchical Reinforcement Learning' at IJCAI 2019.
Visiting student researcher at MILA (summer 2019)
Visiting student researcher at the Montreal Institute for Learning Algorithms (MILA), working with Min Lin and Prof. Yoshua Bengio on Consciousness Prior based on Transformer architectures.
Completed BSc at Tsinghua University
Graduated from Tsinghua University (undergraduate degree completed in 2020).
Started MS at Carnegie Mellon University
Enrolled as an MS student at Carnegie Mellon University (2020–2022).
Ended TSAIL research appointment
Concluded research stint at TSAIL (March 2018 to June 2020) working on RL algorithms and environments.
EnvPool technical milestone: Atari & MuJoCo training speedups
EnvPool demonstrated ability to train Atari Pong and MuJoCo Ant in minutes on consumer hardware, cited as a major systems improvement for RL experimentation.
Joined Sea AI Lab as research engineer (Singapore)
Full-time research engineer position at Sea AI Lab, working with Min Lin and Prof. Shuicheng Yan (May 2021 – Sept 2021).
Created EnvPool (initial release) — ultrafast vectorized RL environments
Developed EnvPool (ultrafast RL environment executor) at Sea AI Lab; released on GitHub; allows training Atari Pong / MuJoCo Ant in minutes on a laptop.
Left Sea AI Lab (end of summer role)
Concluded summer/full-time role at Sea AI Lab (ended September 2021).
Published: Deep RL with Credit Assignment for Combinatorial Optimization
Co-author on paper in Pattern Recognition (Volume 124, 2022) on deep RL with credit assignment for combinatorial optimization.
EnvPool GitHub release & open‑source adoption milestone
EnvPool released on GitHub and gained adoption as an ultrafast environment execution engine for RL experiments (benchmarked in publications).
Tianshou JMLR / GitHub adoption milestone
Tianshou published in JMLR and used widely as a modular PyTorch RL library; core contributor role recognized in academic publication.
Tianshou paper published (JMLR)
Co-authored 'Tianshou: A Highly Modularized Deep Reinforcement Learning Library' in Journal of Machine Learning Research (JMLR 23 (267)).
Completed MS at Carnegie Mellon University
Finished MS studies at Carnegie Mellon University (2020–2022).
Joined OpenAI — Research Engineer, Post‑Training team
Joined OpenAI (Post‑Training team) as a research engineer; responsible for designing and scaling RLHF, post‑training infra and RL pipelines (since July 2022).
EnvPool paper published at NeurIPS 2022
Published 'EnvPool: A Highly Parallel Reinforcement Learning Environment Execution Engine' at NeurIPS 2022 (paper published Sept 17, 2022).
Contributed to ChatGPT initial release
Listed as contributor (6th author) to OpenAI/ChatGPT initial release effort (ChatGPT launched Nov 2022); contributed RL/infra effort supporting release.
Core RL infra author for RFT (reinforcement fine‑tuning)
Listed as core RL infra author for RFT — reinforcement fine‑tuning infrastructure used in OpenAI post‑training workflows.
Published additional RL systems work and open-source contributions
Ongoing development and open-source maintenance of Tianshou and EnvPool; contributed to ecosystem benchmarking and tools.
Contributed to GPT‑4 (RL infra author)
Served as an RL infrastructure author for GPT‑4 (public release of GPT‑4 in March 2023); contributed to RL/post‑training infra for the model.
Contributed to GPT‑4 Turbo and post‑training efforts
Listed in CV as contributing to GPT‑4 Turbo and later o‑series/incremental post‑training releases (post‑training infra).
ICLR 2024 poster: Cleanba distributed RL platform
Co-authored 'Cleanba: A Reproducible and Efficient Distributed Reinforcement Learning Platform' (ICLR 2024 poster).
NeurIPS 2024 submission: Open RL Benchmark (tracked experiments)
Co-authored 'Open RL Benchmark: Comprehensive Tracked Experiments for Reinforcement Learning' (NeurIPS 2024 submission / track Datasets and Benchmarks).
Contributed to multimodal and o‑series model infra
Worked on multimodal RL and o‑series (optimization/operational) post‑training infrastructure as listed on CV updates.
Contributed to GPT‑4V multimodal RL
CV lists contribution to GPT‑4V (multimodal RL), indicating work on multimodal reinforcement learning infrastructure.
Listed contributions to GPT‑4o, GPT‑4.5, RFT, Operator, GPT‑5
CV (updated) lists leadership/authoring roles across multiple post‑training projects and model releases (GPT‑4o infra lead, GPT‑4.5, RFT, early Operator effort, and GPT‑5 post‑training infra). Dates reflect work up to 2025.
CV / public profiles updated listing broad set of model contributions
Public CV and OpenReview profile reflect contributions to ChatGPT, GPT‑4 family models, EnvPool, Tianshou, and multiple research publications through 2024–2025.
Key Achievement Ages
Explore what Jiayi Weng and others achieved at these notable ages:
Similar Trajectories
Carlos Alcaraz
Born 2003 · Age 22
Spanish professional tennis player; multi‑Grand Slam champion; youngest ATP world No.1 (Sept 12, 2022).
Trinity Moyer-Rodman
Born 2002 · Age 23
American professional soccer forward for Washington Spirit (NWSL) and the United States women's national team. Youngest player ever drafted in NWSL (2021). NWSL Rookie of the Year (2021), U.S. Soccer Young Female Player of the Year (2021), Olympic gold medalist (2024).
Shedeur Sanders
Born 2002 · Age 23
American professional football quarterback. Played at Jackson State and Colorado; winner of Jerry Rice Award, Deacon Jones Trophy, Johnny Unitas Golden Arm Award; selected by the Cleveland Browns in 2025 NFL Draft.
Caitlin Clark
Born 2002 · Age 23
American basketball player; NCAA Division I all-time leading scorer, two-time national player of the year at Iowa, 2024 No. 1 WNBA draft pick by the Indiana Fever, 2024 WNBA Rookie of the Year, widely credited with boosting popularity of women's basketball ('Caitlin Clark effect').
Smosh
Born 2002 · Age 23
Smosh is an American YouTube sketch-comedy collective and production brand founded as smosh.com by Anthony Padilla (joined by Ian Hecox). It expanded into multiple channels, products, and ventures, underwent several ownership changes, and was repurchased by its co-founders in 2023.
Ben Shelton
Born 2002 · Age 23
American professional tennis player; left-handed power server; NCAA singles champion turned ATP top-10 player with multiple tour titles including the 2025 Canadian Open (Masters 1000). Coached by his father Bryan Shelton.