Aleksei Prikhodko
Senior AI engineer at Artificial Agency — Edmonton, Canada
Four years shipping LLM agents into AAA games at Inworld: long-term memory, behaviour engines, and the observability needed to trust them under a frame budget. Before that, inverse and ill-posed problems at Novosibirsk State University — three papers, 41 citations on the epidemic-modelling one.
Writingall →
01 Inference optimization decoded: KV cache, Flash Attention, and why your LLM is slower than it should be
The optimizations every team names but few can order by payoff. Walked through in the sequence they actually matter, with the cost model behind each one.
Builtall →
ML Canon59 landmark papers from McCulloch–Pitts to GRPO.
Game Dev Canon12 modules, 54 notes, 15 interactive labs — from arcade hardware to LLM-driven NPCs.
VectorAPLife as eight vectors instead of a to-do list.