Qingyang Wu is a Research Scientist at Together AI (and a Columbia University PhD) who has been a contributor to high-impact open-source RL post-training projects including DeepSWE (a state-of-the-art open-weight coding agent, which he co-led) and DeepCoder, as well as research on accelerating RL training via distribution-aware speculative decoding (an oral presentation at MLSys 2026). Attendees of his session can expect practical, systems-level insights on making reinforcement learning for LLMs faster and more scalable.

Public activity researched automatically · as of Jun 2026