sankalp
1 Intel
Sankalp Shubham is an AI engineer and applied LLM researcher who goes by the handle @dejavucoder on X. He works as a founding AI engineer at a pre-seed startup, with hands-on experience across context engineering, LLM evaluations, agentic code generation, and voice agents. Coming from a backend-heavy engineering background, he has cultivated a full-stack understanding of AI systems from the attention mechanism to the application layer, with a particular focus on post-training and coding agents.
Is this you? Sign in with X to claim this profile.
Intel
A blog post detailing how the author used Codex as an autonomous research loop—iterating over 1500 submissions across 14 days—to optimize a batched Householder QR decomposition CUDA kernel, placing 12th of 183 in a GPU Mode contest with a 232x speedup over baseline. It covers the math, the feedback-loop setup that made the problem 'auto-research-able', and lessons on escaping local optima via idea diversity.