Rajshekar RC

Full Stack Developer Intern · Bangalore, Karnataka

• Engineered a production-ready e-commerce platform with product catalog, user authentication, cart management, and order processing using React, Express, and MongoDB — delivering 5+ core features end to end.
• Designed and optimized 10+ RESTful backend APIs improving database query efficiency and eliminating redundant server-side operations across the application.
• Built a fully responsive mobile-first UI across 8+ pages ensuring seamless and consistent UX across all screen sizes and devices.

Production multi-agent research system · Live Demo

• Shipped a 5-agent LangGraph research system (Planner, Searcher, FactChecker, Writer, Critic) with a Critic→Writer revision loop, deployed across Render (FastAPI) and Vercel (Next.js) with Groq→Gemini failover and live health monitoring.
• Delivered 35% F1 gain (0.055→0.074) over RAG baseline on HotpotQA, cut hallucination 41%→22% via fact-checking, and trimmed multi-hop latency ~40% via parallel asyncio.gather search; validated by a 134-test pytest suite (4s CI).

Production AI resume analyzer — Live Demo

• Shipped a production-grade AI resume analyzer in 15 days with 5 LLM-powered endpoints — ATS scoring, JD matching, keyword extraction, bullet rewriting, and a 2-stage RAG pipeline indexing 53 real job descriptions via ChromaDB and Llama 3.3 70B on Groq (25× latency gain, avg 1.25s).
• Engineered structured output reliability using Pydantic v2 schema validation with auto-retry logic, achieving 100% schema validity across 15 eval test cases and verified prompt injection defense against adversarial resume inputs.

Medical ML classification system

• Built a medical ML classification system with full pipeline across 5+ stages: data preprocessing, feature engineering, model training, and accuracy benchmarking — achieving measurable improvement over baseline using production-level data science practices.

Real-time eye-tracking system

• Developed a real-time iris localization and eye-tracking system using MediaPipe and OpenCV with facial landmark extraction, processing 30+ frames per second for smooth low-latency live-feed performance.

Bachelor of Engineering in Computer Science · Shivamogga, Karnataka