He founded ReWoo — an AI-workforce startup that deploys autonomous agents to run entire business functions for companies. First paying client in 10 days. That client was in the UK.
Before ReWoo, he built and ran Toolly — an AI tools discovery platform with 400+ curated tools, a community submission pipeline, and a Learn AI hub. Built solo. Maintained for 11 months. Then open-sourced entirely.
He is the first author of a published research paper on Bengali speaker diarization, presented at BUET CSE Fest 2026, achieving 56% inference speedup — making speech AI accessible to 230 million speakers who had almost nothing built for them. He is a Kaggle Master, ranked in the top 1% globally across 4,082 competitors.
He currently serves as AI Engineer and CTO on the founding team of a UK-based fintech startup — building production Finance AI agents that handle real money and real consequences.
Build it. Ship it. Make it earn.
Every system he touches must be reliable, observable, and defensible under real constraints. Not impressive in a notebook. Useful in the world.
CAREER
ReWoo
Founded ReWoo, an AI startup engineering autonomous agentic workflows and multi-step reasoning systems. Leading architecture design for production LLM orchestration, stateful agent graphs (LangGraph), and enterprise AI automations.
UK-based Startup (Remote)
BUET CSE Fest 2026
Bangla Diarizz: Domain-Adapted Bengali Speaker Diarization via Knowledge Distillation. Achieved state-of-the-art DER 0.19 (dev) / 0.286 (private LB) with a distilled student model delivering a 56% inference speedup and running 3.4× real-time on CPU for resource-constrained deployments. Preprint on ResearchGate →
Toolly
Architected, launched, and scaled toolly.tech — an AI discovery platform featuring 400+ curated tools across 15 categories, an automated moderation pipeline, and an integrated Learn AI hub. Built solo across 200+ commits; engineered Toolly Studio (Dockerized image generation with Streamlit & Bria AI).
Project Selection
A curated collection of systems I've built. These projects are selected to demonstrate end-to-end engineering—from custom deep learning architectures (Transformers from scratch) and agentic AI middleware, to production pipelines that solve real business constraints like latency, scalability, and vendor lock-in.
Research
A lightweight pipeline for Bengali long-form audio that reaches DER 0.19 (dev) / 0.286 (private LB), with a distilled student model that runs at 3.4× real-time on CPU and roughly 56% faster inference than the baseline — aimed at deployments without heavy GPU infrastructure.
Preprint on ResearchGateCurriculum
A structured, project-driven curriculum of production-grade AI systems. Each project is numbered, open-source, and designed to take you from zero to shipping real AI in the real world.
| # | Project | Description | Tags |
|---|---|---|---|
| 101 | LLM Playground | Production-level large language model training and deployment framework | LLMTrainingMLOps |
| 102 | Customer Support Chatbot | Production-ready AI customer support with RAG, hybrid search, reranking, and PEFT fine-tuning | RAGProductionLLM |
| 103 | Ask the Web | Perplexity-like AI agent with ReACT/ReWOO/Reflexion reasoning, MCP & A2A protocols, and multi-agent orchestration | AgentsSearchRAG |
| 104 | Deep Research | Deep research AI using web search, OpenAI o3, and DeepSeek-R1 with inference-time scaling | ResearchLLMAgents |
| 105 | Multi-Modal Generation | Production T2I and T2V synthesis with full VAE, GAN, DiT, and DDPM/DDIM/DPM-Solver++ implementation | T2IT2VDiffusionDL |
Right now
Focusing on my own AI startup, ReWoo, while also serving as CTO for a UK-based startup.
Containerized ML services on AWS with FastAPI, Docker, and CI/CD — production inference that handles real traffic, not just localhost demos.
Agentic workflows with LangGraph — multi-step orchestration, tool calling, and the evaluation challenges that come with autonomous AI systems.
Open to AI/ML engineer roles, production ML consulting, and research collaborations where the bar is shipping, not slides.
Kaggle Master — Top 1% (29/4,082) in Road Accident Risk. 30 competitions total. View Kaggle profile
CONTACT
I'm open to AI/ML engineer roles, production ML consulting, and research collaborations. If you want a system that actually ships — reach out.