ai-ml
Masaar
AI Adaptive Assessment Platform — FastAPI, LangGraph, LiteLLM, Celery, E2B, Qdrant
FastAPILangGraphLiteLLMCeleryE2BQdrant
Jun 2026 – Jun 2026GitHub
Highlights
- Engineered an agentic assessment platform where each evaluation runs as a single adaptive LangGraph-orchestrated chat session, combining LLM-driven challenge generation with live code execution.
- Owned the E2B code-execution vertical end-to-end: built multi-language sandboxed runners, an adaptive submit flow, and Qdrant-backed memory cards for cross-session recall, integrated with the examiner agent.
- Shipped cross-cutting reliability work — API rate limiting, Celery task fallbacks, and an admin judge-review UI — hardening the platform for production use.
- Designed a five-dimension competency radar report (Thinking, Soft Skills, Work, Digital/AI, Growth) delivered to candidates post-assessment.