Ayan Aditya

I've built three complete systems from first principles: a multi-threaded C++ packet inspection engine, a vector database with a working RAG pipeline, and an encrypted Java Spring Boot payment backend.

For my DPI engine, I engineered dynamic thread-pool load balancing and 5-tuple session reconstruction, sustaining 10,000 packets per second at sub-2 ms latency. I also deployed SNI-based domain filtering that reduced false positives by nearly 30% in real-time blocking scenarios.

I built VectorDB in C++ with HNSW, KD-Tree, and brute-force search, exposing the system through a REST API and benchmarking interface. The project includes an end-to-end RAG workflow that chunks documents, retrieves 768-dimensional embeddings through a custom HNSW index, and sends grounded context to a local LLM.

For UPI Offline Mesh, I designed secure offline payment settlement over an untrusted device mesh using RSA-2048 and AES-256-GCM. I've also completed JPMorgan Chase & Co.'s software engineering virtual experience, building a real-time stock data feed and trading-position dashboard with Python, REST APIs, and Perspective.