I am a systems and ML infrastructure engineer obsessed with making LLM-powered tools performant, predictable, and reliable at scale.
Featured Project:
Eulix: Built a production-grade code intelligence system (Rust parser, Python embedder, Go query engine). It parses 26M lines of code into ASTs and custom call graphs in less than 60 seconds, serving sub-300ms multi-level retrieval queries at 1.5x RAM (~2GB) with 98% retrieval accuracy.
What I bring:
Deep Infrastructure Optimization: Experienced in designing custom binary formats, optimizing inference serving pipelines, and debugging platform-specific runtime issues (e.g., ROCm/HIP on AMD hardware).
Open Source Contributor: Contributed non-obvious bug fixes to major upstream projects like Gin-Gonic and Canonical/Pebble.
Other Projects: Built Vanish (Go-based modern rm alternative with a TUI) and Tyr (Rust-based ML file organizer).
Looking to join an engineering-first team tackling hard infrastructure problems at the intersection of systems programming and AI.
I just read about system designs/agorithms for an hour max and take small notes helps me stay upto date without having a information overload.
Regrading the too much raw info issue is real and I still struggle with it but i have learned to ignore it till i actually need that info
Technologies: Rust, Golang, Python, C, shell, Linux Kernel,
MYSQL/MariaDB
Something About me: I believe in fast and secure code best example where I showcased it on eulix(recent OSS project), its a code analysis tool to speed up onboarding of devs and understanding legacy code by using parsed ast+embeddings+source code, right now it can parse linux kernel source code 26m loc in just a min retrival accuracy is 90+ and answer accuracy depends upon llm used.
when my friend cloned my voice rvc or something model from github and was creating bad songs, it was funny but GOD DAMN i got called into HoDs office for that