Rahat Kabir
I build AI agents and developer tools — and write about what breaks when you ship them.
Always shipping — agents, CLIs, infra experiments.
Technical writing
Essays on AI engineering, system design, and lessons from production.
-
Measuring Cold Starts on a 27B Hugging Face Endpoint
Oct 1, 20268 min readI deployed Qwen3.8-27B on a dedicated Hugging Face Inference Endpoint, scaled it to zero three times, and measured what cold starts actually cost: 6.6–10 minute boots, 36–50 hard 503s, and roughly $0.28–$0.42 of billed startup per wake.
-
I built an AI audio briefing app and moved it to AWS
Jul 2, 20269 min readHow I moved briefcast from a local AI briefing app into a production-style AWS setup with queues, workers, storage, observability, and cost trade-offs — and then imported the whole stack into Terraform.
-
I built my own AI coding agent. Then I actually used it.
Jun 10, 20269 min readA week of dogfooding astra-claw, my CLI coding agent: emergent OCR pipelines, silent package installs, memory that forgets its own lessons, and approval fatigue.
Work
Open-source tools and products I've designed, built, and shipped.
Astra-Claw
A terminal AI coding agent in Python — tool loop, persistent memory, markdown skills, and an approve-before-edit gate.
View project →
Job Search Agent
AI-powered job search assistant built with LangChain DeepAgents that reads your CV, searches the web for matching jobs, and ranks them by fit -- all through a chat interface with human-in-the-loop approval.
View project →-
Briefly AI
Open Source Maintained 2026Python CLI for creating concise AI briefs from text, local files, URLs, YouTube videos, and PDFs.
-
Promptvault
Product Live 2026Personal prompt library — save, organize, and run your best AI prompts in one place.
-
Data Analyst
Product Soon 2026Conversational data analyst for CSV/Excel files with descriptive analysis, interactive charts, dashboards, and reports.
-
DELEGATE
Product Soon 2026AI-powered travel agent for flights, hotels, events, and real-time news search with voice and text chat.
About
I'm an AI engineer based in Dhaka, Bangladesh, working where LLMs meet real systems — deploying agents, debugging them when they fail, and shipping them past the demo. The fastest way to reach me is email; GitHub has the latest of what I'm building.