DEV Community

Rafia Minhaj
Rafia Minhaj

Posted on

TechPulse AI — Autonomous Engineering Interview Errata & Verification Agent

#devchallenge #sanitychallenge #sanity #ai

Sanity Challenge Path One Submission

This is a submission for the Sanity Challenge, Path One: Ship an Agent That Queries Real Content

What I Built

I built TechPulse AI, an autonomous engineering interview verification agent designed to resolve contradictions between company coding specifications, rulebooks, and edge-case errata for top technical interview drives (such as Booking Holdings, Infosys SP, IBM, and AWS).

TechPulse AI queries structured JSON documents stored in a Sanity Content Lake via the Sanity Context MCP Endpoint, evaluating claims side-by-side with source provenance and returning authoritative, source-verified answers.

Demo

Code

The project architecture consists of:

  1. Sanity Content Lake Schema: interviewQuestion, companySpec, errataDoc.
  2. FastAPI Microservice: Async agent backend connecting to Sanity Context MCP.
  3. React/Next.js Dashboard: Interactive query interface displaying source claims side-by-side.

How I Used Sanity

I modeled technical interview knowledge into structured JSON schemas inside Sanity:

  • companySpec: Tracks time limits, memory limits, and platform constraints.
  • errataDoc: Stores verified edge-case clarifications and rule contradictions.

Key Sanity Context Tools Used:

  1. Sanity Context MCP Endpoint: Exposed structured datasets through MCP to our Python AI agent.
  2. GROQ Queries: Retrieved precise schema relationships linking errata back to authoritative company specs.
  3. Knowledge Base Indexing: Distilled technical docs into indexed, queryable chunks where every claim retains source provenance.

When a query is asked, TechPulse AI fetches both the rulebook spec and the errata doc simultaneously, surfacing conflicting constraints side-by-side with direct links to the underlying Sanity documents.

Sanity Project Details

  • Sanity Project ID: techpulse_ai_2026
  • Dataset: production

Agent Session

  • Agent Session Transcript: Embedded via Gemini CLI and FastAPI MCP client.

Author

Rafia Minhaj

B.Tech CSE '27 | GSSoC '26 Global Rank #29 | Microsoft Certified AI Agent Engineer

GitHub: https://github.com/rafiaminhaj

Top comments (0)