Software Engineer · systems, ML, full-stack
I like building things that are fast, private, and useful.
systems → compilers, distributed infra, HPC (MPI / OpenMP / SIMD), low-level
ml / ai → PyTorch, MLX, transformers, generative models, RAG pipelines
full-stack → React, Vue/Nuxt, Next.js, FastAPI, real-time apps
security → RSA / AES / PKI, E2E encryption, privacy-first architecture
| project | what it does |
|---|---|
| montage-ai | AI video pipeline — LangGraph multi-agent, Gemini Flash, FastAPI, React |
| crag-bench-rag | 6 RAG strategies benchmarked on noisy web corpora (HyDE, Fusion, CRAG…) |
| Zeta | Readable programming language — lexer, scope-aware symbol tables, regex engine with automata |
| chattr | Real-time E2E encrypted chat — RSA key exchange, Nuxt 3, Supabase, delivery receipts |
| ipfs-dht-simulator | Ring DHT with O(log N) routing tables and B-tree local indexing |
| diffusion-transformer | SiT-B implementation for diffusion-based image generation |
| multimodal-rag | Multimodal retrieval — CLIP embeddings, FAISS search, MLX-powered Mistral LLM |
| cyclegan-face-sketch | Bidirectional face↔sketch translation — CycleGAN + Flask UI |
| pacman-sfml | Multithreaded Pac-Man — pthreads, mutexes, semaphores |
| semantic-search-module | Embedding strategies benchmarked across FAISS & Chroma, visualized in Streamlit |
| nlp-urdu-poetry-mlx | Urdu poetry generation — RNN / LSTM / Transformer trained with MLX |
| pulselog | Uptime monitor — Laravel 12, scheduled health checks, public status pages |
languages = ["Python", "TypeScript", "C++", "PHP", "Swift", "C", "Assembly"]
web = ["React", "Vue/Nuxt", "Next.js", "Laravel", "FastAPI", "Express"]
data = ["PostgreSQL", "pgvector", "MongoDB", "Redis", "Supabase", "FAISS"]
ml = ["PyTorch", "MLX", "LangGraph", "LangChain", "HuggingFace"]
infra = ["Docker", "Vercel", "Fly.io", "Railway"]
hpc = ["MPI", "OpenMP", "SIMD", "pthreads"]
also = ["WebSockets", "OAuth", "System Design", "RAG", "PKI"]