A LeetCode-style arena where people solve, submit, and compete — sandboxed multi-language code execution, live contests, curated DSA sheets, discussion, and AI code review, all in one product.

• ⚔️ In-browser judge — Monaco editor + sandboxed execution against test cases with real verdicts
• 🏆 Contests & leaderboards — timed rounds with per-contest and global rankings, live over WebSockets
• 📚 Structured DSA sheets — day-by-day, phase-ordered practice tracks with progress dashboards
• 💬 Community & AI review — discuss threads, solutions, and AI-assisted code feedback
• 🛠️ Admin authoring — full UI to author problems, testcases, and contests
• ☁️ Dockerized and deployed on AWS Lightsail behind Caddy

Type a topic, get a full, editable slide deck — an async AI pipeline that writes the content, generates the imagery, and finalizes a polished, exportable presentation.

• 🧠 Prompt → deck — LLM-driven outline, copy, and layout generation
• ⚙️ Multi-stage worker pipeline — Redis-backed content → image → finalize jobs, so generation scales and never blocks the UI
• 🖼️ Live editor — resizable panels and carousels to refine slides, with PPTX export
• 🔐 Auth, billing (Razorpay) and observability baked in
• 🏗️ Built on TanStack Start (SSR React) with type-safe routing and data

Predict → Plan → Talk. A rank-to-seat guide for the JEE/JoSAA counselling window — engineered to sit near $0 when idle and absorb lakhs of students in minutes when a round result drops.

• 🎯 College predictor over 11,261 official JoSAA 2024 cutoffs across 121 institutes (IITs + NITs + IIITs + GFTIs), with home-state quota applied automatically
• 📋 "List Doctor" — a server-authoritative choice-list checker that flags the classic ordering mistakes before you lock in
• 👥 Mentor marketplace — book paid 1:1 video calls with verified seniors
• 🌩️ Serverless AWS (Lambda · API Gateway · DynamoDB · CloudFront · Cognito) via AWS CDK, architected to survive a 100× traffic cliff
• 🧩 ~70 screens across student / mentor / admin, split into 9 bounded contexts

A voice-driven interviewer that runs a candidate through a real loop — Coding, Technical, and HR rounds — tailored to a job description, with questions grounded in real past interview questions, live proctoring, and a rubric scorecard.

• 🗣️ Realtime voice — streaming STT → LLM → TTS over WebRTC (LiveKit Agents), fully swappable stages
• 🔎 Source-grounded questions — agentic RAG (pgvector/Qdrant + PageIndex + rerank via LangGraph) pulls the questions actually asked for that company/role, then twists them harder
• 👁️ Client-side proctoring — MediaPipe face/gaze detection + tab-focus & paste signals, streamed to an integrity timeline
• ⚙️ Live code execution in the coding round + an LLM rubric scorecard with specific, cited feedback
• 💡 Built cost-first: self-hosted Whisper/Piper now, premium providers later with a one-line swap