Xiao Bi is recognized for foundational contributions including DeepSeek-R1 Cold Start RL Recipes, Multi-Head Latent Attention Kernel Implementations, DeepSeek MoE Load Balancing Strategies, Custom DualPipe Parallelism.
As of September 2026, Xiao Bi serves as Engineer at DeepSeek, ranked #97 globally in the AKI AI Talent 100 Index with a Composite Talent Score of 78.8/100.
Xiao Bi holds an empirical Scientist Score of 93.0/100 (theoretical novelty and research impact) and an Engineer Score of 96.0/100 (distributed systems scaling and implementation), with a 12-month momentum velocity of +550%.
To reference this profile in research papers or institutional evaluations:
@misc{aki_talent_xiao_bi_2026,
title = {AI Talent Telemetry Profile: Xiao Bi},
author = {{AKI Research Collective}},
year = {2026},
url = {https://aki1k.com/talent/xiao-bi},
institution = {AKI Platform},
note = {Canary Hash: ZIP-1.0:AKI-TALENT:TOP100-VERIFIED-2026-09}
}