Tharun Krishna

I’m a full-stack developer and ML engineer who turns AI/LLM capabilities into production-ready systems—prioritizing security, performance, and measurable outcomes. I enjoy building end-to-end products that connect user workflows to reliable back-end intelligence.

Most recently, I worked on a VS Code chatbot that uses a local LLM (Ollama/LLaMA3) to classify security queries and automatically route them to SonarQube, Snyk, or Burp Suite via Model Context Protocol (MCP). I designed a dual-mode scan engine (local filesystem before git push, plus remote SonarQube queries), integrated into a Node.js/Express backend with 18 REST endpoints and SSE streaming.

I also validated the system on a live enterprise SonarQube instance, detecting 24 real security issues across 18 files. That work is being prepared as a research paper for IEEE SecDev / ACM SCORED, focused on MCP-based security tool orchestration.

Earlier experiences strengthened my ML and architecture fundamentals: I researched TimeLLM for volatile financial time-series forecasting, benchmarking against ARIMA and GARCH to improve directional accuracy. I also built a real-time anomaly detection pipeline that reduced false-positive alerts by 35% and deployed inference services on AWS EC2 with Docker for sub-100ms latency.