Siddharth Sharma
Systems Engineer · GPU / HPC · Full-Stack
I build fast, correct systems from TCP stacks in C++ to GPU-accelerated Monte Carlo engines and production LLM pipelines.
Systems Engineer · GPU / HPC · Full-Stack
I build fast, correct systems from TCP stacks in C++ to GPU-accelerated Monte Carlo engines and production LLM pipelines.
Click on the project to open it. →
Built a columnar OLAP engine from scratch with a Vectorized Volcano-model executor. Implements full query pipeline — scan, filter, project, aggregate.
Reimplementation of Cvetanoska & Stojanovski — pricing American call options via Bermudan backward induction. 5 compute backends share a single math core: BS, LCG, Moro Inverse CND, Sobol/Halton + Brownian Bridge.
Complete TCP/IP stack built in both C++23 and Rust, bypassing the OS. Full TCP state machine, connection teardown, retransmission — by-the-RFC.
Personal knowledge base: ingests text/audio/images → semantic chunking → Qdrant vector DB → spaced repetition, RAG chat, live voice via Gemini WebSocket.
Intelligent voice assistant with real-time Speech-to-Text, TTS, synchronized facial animation rendering. Microservices-based architecture using ONNX/TFLite.
Lightweight RAG app running Gemma 3 270M locally. Full PDF parsing, chunking, embedding, and retrieval pipeline with sub-300M param inference.
254 consecutive days of problem solving. Not sprints. Sustained execution.
I'm looking for roles where I can work on hard problems at the systems level whether that's HPC, ML infra, backend systems, or full-stack product engineering.
Thapar University · Punjab, IN · Batch 2027