Based in Dhaka, Bangladesh
AI & Machine Learning Developer
I build systems that see, understand, and act — from edge-computing drone pipelines to cloud-scale RAG architectures. Self-taught, production-focused, and always at the intersection of research and real-world constraints.
Selected Work
A production-grade RAG backend for querying Bengali upazila data in English. Achieves ~4s average response time using ChromaDB and Mistral embeddings, with cross-lingual semantic matching that bridges the language gap cleanly.
A multimodal AI agent that operates a Linux desktop autonomously — optimized for low-spec hardware within 4GB RAM. Uses zero-disk RAM frame captures and kernel-level scancode injection for Wayland navigation.
End-to-end edge-computing pipeline streaming live drone video to an Arch Linux server for real-time inference. Fine-tuned YOLOv8n with dynamic quantization for aggressive speed-accuracy tradeoffs in drone environments.
Capabilities
Languages & Frameworks
AI / Data
Tools & Deployment
Domains
Background
I'm a self-taught AI developer from Dhaka who got into machine learning the hard way — through rigorous independent study in differential calculus, linear algebra, and probability, specifically applied to how neural networks actually work under the hood.
I don't build demos. I build systems that run under real constraints: limited RAM, free-tier cloud, low-power edge hardware. That discipline shapes everything — from how I architect APIs to how I think about model quantization.
Currently looking for an internship where I can contribute to production-grade AI infrastructure, learn from experienced engineers, and push further into agentic and multimodal systems.
Let's connect
I'm actively seeking internship opportunities. If you're working on something interesting in AI, vision, or intelligent systems — reach out.