Autonomous Career & Interview Copilot

Inspiration

Preparing for modern technical interviews is often fragmented, static, and lacking in real-time adaptive feedback. We wanted to build an intelligent, autonomous companion that doesn't just ask static questions, but acts as a true engineering mentor—orchestrating multi-turn workflows and maintaining deep context across sessions.

What it does

Our Autonomous Career & Interview Copilot is built as a self-hosted Model Context Protocol (MCP) server (conforming to spec version 2025-11-25 over Streamable HTTP) integrated with Amazon Bedrock. It ingests user resumes and target job descriptions, automatically constructs personalized multi-day technical interview roadmaps, and evaluates user responses in real time using agentic reasoning.

How we built it

Core Protocol: Built using the Model Context Protocol (MCP) server architecture with Streamable HTTP transport.
AI & Agentic Reasoning: Powered by Amazon Bedrock for dynamic prompt execution, tool orchestration, and intelligent response evaluation.
Backend Architecture: Modular Python backend with robust error handling and clear setup workflows.

Challenges we ran into

Navigating the exact specifications of the MCP Streamable HTTP transport layer and tuning Amazon Bedrock tool-calling schemas required careful iteration and debugging.

What we learned

We gained deep expertise in building production-grade MCP servers, managing multi-turn agent state, and leveraging AWS developer tools for autonomous workflows.