S.YOGADHARSHAN

Rajapalayam

• Architected and solely built RIT Canvas (ritcanvas.in), the college LMS, and its React Native Android app: 55-table schema, role-based access, 2,139 students and 169 educators.
• Engineered backend services for AI quiz generation from PDF, DOCX and PPTX (Docling, Ollama), a real-time quiz engine with server-side timers, and cron jobs.
• Run the platform fully self-hosted, not on cloud hosting: 10 Docker replicas deployed with Dokploy behind Traefik and a Cloudflare Tunnel, plus a self-hosted Convex backend.
• Co-developed the RIT ERP (ritrjpm.edu.in) in a 7-member team on Python, Django and MySQL: 14 production modules, including student and staff portals and online fee payment.
• Built RIT Pulse, a live quiz platform with server-scheduled timers designed to save about 90,000 database reads per question at 300 players, and co-built RIT REACH's proctored exams.

solarscript.in

• Build and sell 10+ FiveM (GTA V) scripts in Lua, React and MySQL for ESX, QBCore and Qbox; 50+ custom scripts delivered.
• Designed a 35-table MySQL schema for a 17-page police dispatch and records system with 151 client-server callbacks.
• Modelled 3D props in Blender (in-car screen, surveillance kit, house shell) and shipped them as native GTA V assets.

• Co-organised RIT's 24-hour hackathon: 31 teams, 114 participants, 50 problem statements and a Rs. 30,000 prize pool.
• Built the event website (hackfinity3.0.ramcoad.com) and the React Native judging app with role-based access, a configurable 10-criterion rubric and QR check-in.

• Built semantic search with sentence-transformer embeddings, structured LLM output and LangChain pipelines in Python.

AI Video Clipping Agent Python, YOLO, faster-whisper, FFmpeg

• Built a six-stage pipeline that cuts long videos into captioned vertical shorts in Tamil, Hindi and Telugu, with parallel GPT and Gemini planning and YOLO face tracking.

LUMEN query agent Python, FastAPI, LangGraph, PostgreSQL, MongoDB

• Designed a five-step agent that writes, validates and explains SQL and MongoDB queries from plain English, blocking destructive statements.