I’m currently in the R&D team at NCAI, working on multimodal retrieval, text-to-SQL agents, and LLM training. Recently we published two papers, LARE and MINER.
Before that, I worked at TAHAKOM with the NLP team on LLM post-training, and we got first place on the Open Arabic LLM Leaderboard at the time. I also worked on agents there.
Before that, I worked at two startups, Mawj doing speech stuff and Muragib for social listening, and at Microsoft on vision and telemetry.
Recently I’m doing some stuff in distributed training, parallelism and GPU numerical linear algebra, mostly in Julia. Some of my contributions: CUDA.jl, GPUArrays.jl and Dagger.jl.
Besides that, I’m into pop culture.
