Sayan Das

A full-stack fake news detection system analysing articles, claims, screenshots, and social posts in any language, returning an explainable authenticity score backed by a Chrome extension for one-click verification.

• Adaptive two-path scoring engine fusing NLP, ML classification (RoBERTa, LR/TF-IDF), Groq LLaMA-3.3-70b semantic analysis, and fact-verification via asyncio.gather().
• Multimodal screenshot/social-post analysis using a vision model for claim extraction, plus multilingual support via automatic language detection and NLLB-200 translation.

An AI agent that evaluates every new GitHub issue against a team's uploaded PRD through three sequential analysis passes and — for compliant issues — automatically generates production-ready code changes and opens a pull request, all with zero human intervention.

• Three-pass AI evaluation pipeline (Drift Detection, Strategic Alignment Review, Rule Enforcement) returning structured JSON from a single Gemini 2.5 Flash call.
• Event-driven backend with Inngest for durable, retry-safe background processing of GitHub webhooks (HMAC-verified) and Octokit for automated comment posting and issue labelling.

A full-stack sports management platform consolidating stat tracking, club/tournament organisation, and live ball-by-ball/event scoring for Football, Cricket, and four other sports into a single dashboard for amateur and semi-pro communities.

• Multi-sport data model (Prisma + PostgreSQL on Supabase) supporting per-user, per-sport profiles with independent stat tracking and automatic stat sync on match completion.
• Real-time live match scoring for Football (goals, cards, substitutions, penalty shootouts) and Cricket (ball-by-ball events, dismissals, extras, bowling figures) with idempotent stat-sync transactions.