CS student turning hackathon weekends into real products.
I build AI-first apps that ship β RAG pipelines, full-stack TypeScript, and ML that runs in production.
Most of what's below is deployed and usable right now.
Five of the projects below β Academic Twin, CivicSetu, MedWatch, SpendWise, and Student Risk Analytics β are part of Veltron Group, an umbrella I'm building for AI products that actually ship.
Academic Twin β an AI study companion that reads your course material.
Upload your PDFs and notes, then chat with them, generate quizzes and flashcards, and get a study plan built around your weak topics. Retrieval-augmented, so every answer cites the source it came from.
Next.js 14 Β· Prisma Β· Pinecone Β· LangChain Β· Clerk Β· Supabase β live demo β
ποΈ CivicSetuReport. Route. Resolve. Most civic apps stop at the complaint. CivicSetu treats it as the start β reports are clustered and routed to whoever can act, including universities and industry who take the real ones on as projects. Eight stakeholder roles, bilingual, installable PWA with AI triage.
|
π DataSage AIChat with your data. Upload a CSV or Excel file and get an instant auto-generated dashboard β KPIs, six chart types, an AI-written insights panel, grounded chat, and a one-click executive PDF report.
|
π Kalvi.AITamil-first AI education. An education platform built for students and schools in Tamil Nadu β designed around the local language first rather than bolting translation on afterwards.
|
Catching students before they fall behind. ML platform that predicts academic risk from performance data and surfaces it to institutions through real-time dashboards.
|
Also: MedWatch β a regional medicine-shortage early warning dashboard, built for Manipal Hackathon 2026 (SDG 3) Β· live β Β· SpendWise β see the trade-off before you spend, budgeting that shows what an impulse buy actually costs your goals Β· live β Β· CHIC Connect β clinic site with live doctor data, real appointment booking, an authenticated admin panel and server-side email.
Beyond the web stack I work on applied deep learning β multi-task CNNs, Grad-CAM explainability, and model evaluation on real clinical datasets.
Open to internships and collaborations on anything AI-shaped.


