Products and projects I've designed, built and shipped — spanning AI, machine learning, mobile, web, trading automation and security.
A production Flutter app for AI-powered social media content generation — integrating 40+ AI models across providers, with multi-platform campaigns, voice-to-post, a "Brand DNA" personalization system, an agency mode for managing clients, and Razorpay credit-based monetization.
An AI-powered startup idea validator built with Next.js & React on a Firebase backend. Integrates dual Claude + Gemini providers with live web search via Serper for grounded, real-time validation — plus a Razorpay-powered Pro tier with full payment and credit handling.
An automated NIFTY/BANKNIFTY options engine using hedged strategies (Iron Condor, Bull Put / Bear Call spreads). Features a full Black-Scholes Greeks engine, IV-Rank entry signals, real-time Zerodha Kite data over WebSocket, Telegram bot control and one-click Railway deployment with built-in risk controls.
A zero-cost, one-command recon pipeline that turns any IP into a live attack-map in under 6 minutes. Couples a Gemma-3n LLM for AI wordlist generation with SYN-stealth + version scanning, Cloudflare-aware content discovery and tech fingerprinting — surfacing only actionable findings with near-zero false positives.
The ZYRK-Recon engine shipped as a live REST API on RapidAPI. Turn any domain or IP into an actionable attack-surface map — port & version scanning, Cloudflare-aware content discovery, tech fingerprinting and AI-driven wordlists — through one simple endpoint any app can call.
A machine-learning model that predicts cardiac-arrest risk from patient health metrics. Built in a Python / Jupyter workflow with data preprocessing, feature engineering, model training and evaluation.
An ML classifier that detects phishing websites and URLs by analysing their features — trained and evaluated in a Python / Jupyter notebook, blending data science with a security mindset.
A spam-vs-ham email classifier using NLP feature extraction and machine learning, built and evaluated in Python / Jupyter to accurately filter unwanted mail.
A Python tool that analyses legal documents — extracting and summarising key clauses and information with NLP / LLM techniques to make dense legal text far easier to understand.