TECH
RUNS
DEEP.

Everyone uses ChatGPT. Few people understand how it actually works. I built a decoder-only Transformer language model from scratch in PyTorch — no nn.TransformerDecoder, no Hugging Face — to demystify the architecture behind GPT-2. Here is every layer, every matrix multiply, every design decision.

You're staring at five dashboards at 2 AM, trying to figure out why production is down. What if the first hour of investigation could happen automatically? I built an AI Incident Response Platform — here's why, and the architecture behind it.

You type a question, hit Enter, and the answer streams in word by word. But what actually happens behind the scenes? I built an Enterprise AI Assistant from scratch — here is every single step, from keypress to rendered response.

Basic RAG breaks under complex, relational queries and returns dull Markdown walls. Here's how I rewired my AI app into a Supervisor–Worker multi-agent system, fused it with GraphRAG via Neo4j, and replaced static text with streaming Generative UI components.

I built a production-ready Enterprise AI Assistant using Retrieval-Augmented Generation (RAG) and Mistral LLM. Here’s the architecture, tech stack, and real-world use cases behind this scalable AI system.