Build a High-Speed Streaming Langchain Orchestrator

01The challenge

MyPip (a funded Paragon Research Labs startup) is hiring an AI/Backend Founding Engineer for our upcoming winter co-op cycle. We are dealing with complex LLM operations that need to be fast, secure, and production-ready. We do not want developers who just wrap OpenAI's basic API. We need infrastructure-level thinkers. Your task is to build a highly optimized FastAPI endpoint in Python that uses Langchain to orchestrate a user query and streams the response. REQUIREMENTS: The Endpoint: Build a single POST endpoint /ask that takes a JSON payload {"query": "..."}. Orchestration (Langchain): Use Langchain to route the query. If the query asks about "Math", route it to a simple math tool/chain. If it asks about "General", route it to a standard LLM chain. Real-Time Streaming: The endpoint MUST return a Streaming Response (Server-Sent Events). The LLM output must be streamed back chunk-by-chunk asynchronously. Security: Do NOT hardcode any API keys. You must load keys securely from environment variables.

02Required skills

• python
• fastapi
• langchain
• llms
• asyncio

03Strict acceptance criteria

1. 01Uses Langchain for routing and orchestration.
2. 02Endpoint successfully returns a chunked Streaming Response (SSE).
3. 03ZERO exposed API keys in the code (uses os.getenv or .env).
4. 04Includes a valid Loom video explaining the code logic.

04What to submit