Shijun Li

Shijun Li⁠

About

I research reinforcement learning for recommendations and agentic systems.

Experiences

  1. OMACGave an ICML 2026 oral on multi-agent LLMs

    Our OMAC paper was selected as an oral in Seoul, and we released the code alongside it.

  2. Machine Learning Research Intern at NetflixBeat Netflix's re-ranker on key metrics

    I built an LLM reasoning method for cold-start recommendations, fine-tuned with supervised and RL techniques.

  3. Data Science Intern at IntuitBrought goal-conditioned RL to Intuit recs

    It beat the existing baseline in offline tests on Intuit's real data, with minimal added complexity.

  4. Research Intern at Kuaishou TechnologyWrote 3 papers on conversational recs

    I also brought offline RL into Kuaishou's short-video recommender, which serves millions of people.

  5. Research Intern at University of FloridaModeled grain growth as an RL problem

    I turned noisy microstructure images into low-dimensional representations an RL agent could work with.

Follow Shijun for more...

LinkedIn

© 2026 Delphi · Terms · Privacy

Shijun Li

Shijun Li⁠

About

I research reinforcement learning for recommendations and agentic systems.

Experiences

  1. OMACGave an ICML 2026 oral on multi-agent LLMs

    Our OMAC paper was selected as an oral in Seoul, and we released the code alongside it.

  2. Machine Learning Research Intern at NetflixBeat Netflix's re-ranker on key metrics

    I built an LLM reasoning method for cold-start recommendations, fine-tuned with supervised and RL techniques.

  3. Data Science Intern at IntuitBrought goal-conditioned RL to Intuit recs

    It beat the existing baseline in offline tests on Intuit's real data, with minimal added complexity.

  4. Research Intern at Kuaishou TechnologyWrote 3 papers on conversational recs

    I also brought offline RL into Kuaishou's short-video recommender, which serves millions of people.

  5. Research Intern at University of FloridaModeled grain growth as an RL problem

    I turned noisy microstructure images into low-dimensional representations an RL agent could work with.

Follow Shijun for more...

LinkedIn

© 2026 Delphi · Terms · Privacy

By using this service, you agree to the Terms of Service and Privacy Policy.