I specialize in building scalable, production-ready systems that combine deep learning, backend engineering, and autonomous AI agents. My work spans:

• 🚀 Asynchronous Microservices — FastAPI, PostgreSQL, Redis task queues, Docker containerization
• 🤖 Autonomous AI Agents — Multi-step reasoning, RAG pipelines, tool routing, dynamic dependency injection
• 🧠 Deep Learning — CNNs, LSTMs, Stable Diffusion, ControlNet, edge ML
• 🔐 Production Systems — Biometric authentication, secure databases, idempotent operations
• 📊 Data Intelligence — EDA automation, ML inference, real-time processing

I believe in first-principles thinking: building things from the ground up to understand how they work, then optimizing for production.

scalable-async-task-engine — Production-grade async task processing microservice
A battle-tested pattern for handling long-running work (ML inference, ETL, email delivery) without blocking HTTP requests. Perfect for ML applications that need fast API responses + reliable background job execution.

# One-command setup
docker-compose up --build
# API on :8000, Worker running, Postgres + Redis included

GraphMind-RAG-v2.0 — Multi-PDF Autonomous RAG Agent
Local AI platform with:

• Hybrid BM25 + FAISS retrieval
• FlashRank re-ranking
• Dynamic tool routing
• Interactive Cybersecurity Tutor with persistent memory

Zero external APIs required — fully offline, fully private.

bioguard-edge-attendance — High-performance biometric attendance
Real-time facial landmark tracking + Eye Aspect Ratio (EAR) anti-spoofing. SQLite idempotent ledger. Edge-first design (no cloud dependency).

neural-sketch-diffusion — ControlNet-guided image generation
Transform hand-drawn sketches into photorealistic artwork using ControlNet edge conditioning + Stable Diffusion v1.5.

NativeWeather-Agent — Autonomous multi-step weather agent
Framework-free implementation with LLaMA 3.3-70B, Groq, and Open-Meteo REST APIs. Pure Python tool routing, no orchestration overhead.

Histopathology-Lung-CNN — First-principles CNN for lung cancer classification
Native image data streams, regularized training loops, confusion matrix visualization. Built from scratch without high-level frameworks.

stock-price-prediction-lstm — End-to-end time series forecasting
PyTorch LSTM with automated data fetching (yfinance), preprocessing, and dynamic visualizations.

1. First-Principles Thinking — Build systems from scratch before reaching for frameworks
2. Production-Ready — All code is architected for scalability, reliability, and observability
3. Privacy-First — Prefer offline, local solutions when possible
4. Asynchronous by Default — Non-blocking operations for responsive systems
5. End-to-End Ownership — From data ingestion to deployment

• 💼 Backend/ML engineering roles where I can architect scalable systems
• 🤝 Open-source collaboration on production-grade AI/backend projects
• 🔬 Innovation projects combining AI agents, distributed systems, and edge computing
• 📚 Technical discussions on async patterns, RAG architectures, and ML infrastructure

• 💻 Explore my repositories — they're documented and runnable
• 📧 Reach out with questions, ideas, or collaboration opportunities
• ⭐ If a project helps you, star it! (Feedback fuels improvement)