A collection of my technical projects spanning distributed systems, machine learning, web development, and cloud computing. Each project showcases different aspects of my expertise.
Distributed KV store using one-sided RDMA over RoCE targeting sub-millisecond latencies. Reed-Solomon erasure coding and SWIM failure detection.
Fork of vLLM v0.18.0 adding chunk-level KV cache eviction for RAG workloads. Eviction scores combine retrieval relevance, attention signal, and recency.
Compared five feature-level transmission strategies for server-side keyword spotting using PyTorch CNNs and Silero VAD — from raw PCM to learned CNN embeddings.
Peer-to-peer offline payment system using Apple Bonjour — no WiFi or cellular required. ECDH handshake via Apple CryptoKit.
Cross-platform AI assistant for comparing echocardiogram studies, saving $3.4M+ annually in healthcare costs.
Re-architected legacy Java load generator into high-performance Rust system, improving throughput by 400%.
NLP-powered graph generation tool that interprets natural language queries to create data visualizations from SQL databases.
ML pipeline for audio enhancement using dynamic mixing and noise augmentation, increasing model robustness by 2.1x.
Document recommendation model using TF-IDF vectorization with Scikit-learn and Pinecone hosted on GCP.
Comprehensive iOS application that revolutionizes how users discover, track, and rank their favorite movies and TV shows with social features.