Back-office automation
I’ve used LangGraph to automate internal back-office workflows. Client names and implementations are private.
Sakib Sadman Shajib
I build AI and GenAI systems that review code, automate back-office work, and answer questions over company data.
Type a command or choose one below. Try "boson" for an illustrative trace.
Independent product in development
I’m building an AI reviewer that reads and runs code, tests it from different angles, and suggests what to change. It looks at behavior, code quality, and whether the change follows the project’s own standards.
The worker creates a new order each time the same job runs.
The first write succeeds, but the worker loses its acknowledgement. The retry writes a second order.
Use a stable request ID and enforce uniqueness in the database. Add a regression test that runs the same job twice.
Not output from a live Boson run.
I’ve used LangGraph to automate internal back-office workflows. Client names and implementations are private.
Built retrieval-augmented systems with LangGraph and LlamaIndex and deployed agent systems with AWS Bedrock AgentCore. I’m evaluating Google’s Agent Substrate on GKE as a portability option.
I own the MCP server, AI DevKit, and shared services including fundmore-date and SFTP. I lead engineering delivery and work directly with clients to resolve issues.
Added validation and tests to the account confirmation flow. Merged
Proposed request lifecycle logging that records field names and types without raw values. Closed without merge
Built a plugin that hands a code review or a focused task to Google’s Antigravity CLI. I may return to it as Gemini tooling evolves.
Hive is a self-hosted AI workspace. Development is paused while I plan a migration to a community-maintained project.