AI Developer — autonomous agents, ML pipelines, and the real-time systems that feed them.
Working where AI meets cybersecurity.
I build intelligent systems — machine-learning pipelines, autonomous AI agents, and the real-time infrastructure that feeds them. My work sits where AI meets cybersecurity, with a decade's worth of side-quests through full-stack web and blockchain along the way.
- 🧠 AI agents & NLP: LLM agents that plan and act — security co-pilots, multi-agent narrative analysis, RAG pipelines — running on locally-hosted models, not just API calls.
- ⚙️ Systems that ship: production billing with automated payment reconciliation, live market-data scraping, real-time streaming pipelines, explainable AI for medical imaging.
- 📐 Philosophy: clean, modular, well-documented code — and systems that hold up against real data and real users.
| Project | What it is | Stack | Status |
|---|---|---|---|
| Cortex Strike | Autonomous offensive-security co-pilot — an AI agent that plans and runs multi-step penetration tests, chaining a large offensive-security tool library through a single conversation. Runs on a locally-hosted LLM, so nothing about a target leaves the machine. | Python · FastAPI · React/TS · Local LLM · PostgreSQL | 🔒 In active development |
| Echo Shield | AI social narrative analysis platform — a multi-stage NLP pipeline (sentiment, stance, entities, Urdu/Roman Urdu) that detects emerging narratives, with a multi-agent system that drafts counter-narrative responses. All models run locally. | Python · FastAPI · NLP models · React/TS | 🔒 Deployed & in use |
| Current | Production billing platform for a commercial property with 100+ tenants — meter readings in, tiered bills, PDFs, email/SMS dispatch, and automated bank-SMS payment reconciliation out. Money stored as integer paisas throughout. | Next.js · Prisma · PostgreSQL · Redis | 💼 Live in production (client) |
| PSX Portfolio Tracker | Investment tracker for the Pakistan Stock Exchange — live scraped prices (no paid API), lot-level holdings with automatic splits on partial sells, and a monthly allocation planner. | Next.js · Supabase · TypeScript · Recharts | 🌐 Live |
| Explainable DR Detection | Final-year project: deep-learning detection of diabetic retinopathy from retinal scans, with explainable-AI visualizations and plain-language reasoning. | PyTorch · XAI · React | 🎓 Final-year project |
Full write-ups for each project live on the portfolio.
- 🤖 RAG Chatbot Platform — retrieval-augmented chatbot with its own backend and frontend, wired into a document knowledge base. (backend · frontend)
- 🏥 Health Plan Generator — AI healthcare-plan generator with a self-evaluation loop that critiques and improves its own output. (backend · frontend)
- 📚 Novel-to-EPUB Automator — resume-safe scraping that compiles web novels into EPUBs and emails them when done; Dockerized end to end.
- 📱 Zenith Reader — full-stack web-novel reader shipped across web and Android from a single codebase.
- ⛓️ DeFi & Balancer Tooling — real-time mempool trackers and Balancer protocol experiments. (gateway · deploy)
- 📊 Resource Dashboard — GPU/CPU/memory monitoring for the ML workstations behind the heavier AI projects. (repo)
| Discipline | Tools & Technologies |
|---|---|
| AI / ML | PyTorch, TensorFlow, NLP & Transformers, LLM agents (RAG, multi-agent), Ollama / llama.cpp, explainable AI, computer vision & OCR |
| Cybersecurity | Offensive-security tooling, VAPT automation, OSINT analysis, autonomous security agents |
| Web & Full-stack | React, Next.js, TypeScript, Node.js / FastAPI, Redux Toolkit & RTK Query, Tailwind / shadcn/ui |
| Data & Real-time | PostgreSQL / TimescaleDB, ZeroMQ streaming, Supabase, Redis |
| Blockchain / Web3 | Solidity, Ethereum, Web3.js, mempool & DeFi tooling |
| Infrastructure | Docker, CI/CD (GitHub Actions), Nginx, n8n automation |
Building intelligent systems, one commit at a time. · 📫 blackst1rmail@gmail.com