Trading systems · Integration · Messaging
Java / C++ / C# across equities & trading platforms. Strong in enterprise messaging (Kafka/Confluent, IBM MQ, TIBCO EMS/RV) and exchange connectivity. Ex–VP Technology (Bank of America); currently Shell.
Hands-on work in agentic systems, retrieval-augmented LLMs, quantitative ML, and reproducible AI infrastructure—published under github.com/dselva7. These repos complement enterprise integration experience with modern Python-first AI engineering.
LangGraph-style agent graphs with persistent SQLite memory, durable state, and Model Context Protocol (MCP) for decoupled tools—suited for long-running workflows, checkpoints, and human-in-the-loop gates.
Advanced RAG patterns including Self-RAG-style flows, vector persistence, and retrieval pipelines built with LangChain and ChromaDB for grounded, auditable answers over your own documents.
Signal processing, risk-aware modeling, and alpha factor exploration on market-style data—aligned with Udacity’s AI-for-trading curriculum using Pandas and NumPy for reproducible research notebooks.
A structured library of notebooks and modules spanning NLP and transformers, computer vision, classical ML, and deep learning with TensorFlow, PyTorch, and OpenCV—useful as reference implementations and teaching artifacts.
Container images and compose-friendly setups to run notebooks and inference services consistently across laptops and CI—Bash-driven automation for scalable ML deployments without vendor lock-in.
Also on this account: profile README, Cert (credentials), and dselva7.github.io — 8 public repositories in total.
All repos @dselva7