Engineered for throughput, latency, and scale.

A high-concurrency, multi-threaded TCP server implemented from the ground up in modern C++ to parse and serve the HTTP/1.1 protocol. Engineered to eliminate allocation overhead, saturate multi-core hardware, and sustain tens of thousands of concurrent client sockets without degradation.

• Multi-threaded TCP Architecture: Developed a high-performance TCP server in C++ from scratch that handles the HTTP 1.1 protocol, implementing socket reuse and zero-copy header parsers.
• Extreme Load Scalability: Capable of scaling to over 80,000+ Requests per Second (RPS) while sustaining 50,000+ concurrent client connections under rigorous synthetic load testing.
• Custom Thread Pool & Non-blocking I/O: Implemented a lock-free work dispatching thread pool combined with non-blocking I/O event polling to eliminate OS context switching overhead and minimize tail latency.

An end-to-end Retrieval-Augmented Generation (RAG) query engine utilizing FastAPI, engineered to query across complex, unstructured document repositories with contextual precision. Employs a hybrid search strategy that fuses dense vector cosine similarity with sparse lexical indexing.

• End-to-End RAG Engine: Engineered a scalable Retrieval-Augmented Generation engine utilizing FastAPI, enabling semantic questioning and contextual synthesis over extensive unstructured repositories.
• Hybrid Retrieval Fusion: Utilized a hybrid search strategy that combines dense vector cosine similarity and sparse Elasticsearch BM25 for maximizing retrieval precision, yielding a 94% accuracy rate in ground-truth answer generation.
• Optimized Ingestion Pipeline: Structured the RAG data ingestion workflow using LangChain with custom parsing, intelligent noise removal, semantic chunking, and metadata storage, strictly minimizing context loss when embedding text.

A low-latency collaborative Markdown authoring and knowledge management system. Designed for engineering teams requiring real-time document synchronization with deeply nested directory structures and instant conflict-free state broadcasts.

• Real-Time Broadcast Engine: Engineered a collaborative Markdown editor utilizing WebSockets to broadcast delta updates with sub-millisecond propagation latency, enabling smooth multi-user co-editing.
• Hierarchical Document Hierarchy: Built a robust document management system featuring customizable folders and deeply nested subfolders to efficiently organize and scale team documentation trees.
• Reactive Full-Stack Architecture: Paired a lightweight TypeScript Express backend with a reactive Vue.js client for zero-lag live Markdown previews and synchronized tree navigation.