I build production-grade AI systems. I don't just call APIs; I build models from scratch, architect agentic workflows, and deploy scalable full-stack applications.

• 🔹 Agentic AI Specialist: Expert in building non-linear, stateful AI agents using LangGraph.
• 🔹 Deep Technical Roots: I build Transformers from scratch in PyTorch to master the math behind the models.
• 🔹 Full-Stack Ownership: I treat every project like my own company, handling everything from React frontends to FastAPI backends and Cloud infrastructure.
• 🔹 Code Literate: I can read complex codebases like a book, allowing me to debug, optimize, and integrate AI into existing enterprise systems quickly.
• 🔹 Founder's Mindset: Focused on building stable, "production-ready" tools that solve real business problems, not just cool demos.

• Agentic Workflows & Multi-Agent Systems (LangGraph, LangChain)
• Generative AI & LLM Fine-tuning (Unsloth, LoRA, QLoRA)
• Deep Learning Architecture (Custom Transformers, Attention Mechanisms)
• Production MLOps (Docker, AWS, GCP, CI/CD)
• Scalable Backends & APIs (FastAPI, Django, WhatsApp Webhooks)

A production-ready AI Sales Representative built with LangGraph.

• Uses a state machine to handle non-linear conversations (pivoting between pitching and Q&A).
• Integrated RAG via ChromaDB to answer specific business queries.
• Features intent detection to qualify leads and extract contact information.
• Deployed via FastAPI with WhatsApp Webhooks for real-time customer interaction.

A real-time trend analysis platform.

• Automated a 100+ topic dataset generation pipeline using n8n and ChatGPT for fine-tuning.
• Fine-tuned a 1B-parameter model using Unsloth for 4-bit inference optimization.
• Built a full MLOps pipeline on AWS to handle fetching, summarizing, and serving social media insights.

Designed and trained a ~60M parameter transformer model using only PyTorch.

• Manually implemented Causal Self-Attention, Multi-Head Attention, and LayerNorm.
• Focused on understanding gradients and architectural bottlenecks to optimize model performance from the ground up.

A semantic discovery system for books.

• Combines vector search and emotion-aware filtering.
• Deployed as a serverless service on Google Cloud Run with automated CI/CD.

• Agentic AI: LangGraph, LangChain, RAG (ChromaDB)
• Fine-Tuning: Unsloth, LoRA, HuggingFace Transformers
• Computer Vision: OpenCV, Pillow

• Integrations: WhatsApp Webhooks, REST APIs, NGINX

I’m always looking to collaborate on high-impact AI projects or discuss the latest in Agentic workflows.

Always building systems — not just models.