$ cat projects/sevasaathi.md

SevaSaathi v2 — Live Opportunity Discovery

AI agents search the live web for scholarships, exams, jobs, and events.

Sole developerJuly 2026AI/ML

Ask in free text → the system asks tailored follow-up questions → four parallel web-search agents (scholarships, exams, jobs, events) hunt the live web while a verifier/ranker agent scores results. Evolution of the earlier multi-agent SevaSaathi discovery engine (Scout → Extractor → Analyst → Compiler over live crawls).

// problem

Students and rural families miss scholarships and government schemes because discovery depends on knowing where to look.

// impact

Model-generated follow-up questions personalize every search; Gemini with Google Search grounding is primary, with Groq and OpenRouter free-tier fallbacks so it runs at zero cost. No login or database by design — a signed-cookie 3/day rate limit, localStorage persistence, and calendar export keep it simple and privacy-friendly.

4 search agents + verifierGemini Search groundingNo login · no DBZero-cost model fallbacks

// tech-stack

Node.jsExpressGemini APIGroqOpenRouterVercel Serverless