← back to selected work

selected work · 2026

SignalDesk

Automated opportunity pipeline: scrape, classify with Groq, draft replies, ship to WhatsApp.

summary

  • Ingests opportunity feeds from Reddit, Hacker News, Product Hunt, and GitHub Issues on a daily schedule.
  • Scores each lead against a skill profile with Groq, then stores fit, source, and delivery state in MongoDB.
  • Drafts a tailored first reply and cover letter from portfolio context instead of sending a generic template.
  • Delivers the shortlisted batch through an authenticated WhatsApp Web session for quick human approval.

problem

  • Useful leads expire quickly because public boards reward whoever sees, qualifies, and replies first.
  • Manual scanning burns time before fit, budget, timing, and contact context are clear enough to act on.
  • Generic outreach weakens trust, so every draft needs to reflect the role, portfolio evidence, and timing.

approach

  • Adapters normalize each source into Candidate records with role, budget, timing, and contact context.
  • Groq scores fit against a saved skill profile, while MongoDB hashes prevent repeats across every source.
  • A drafting pass uses portfolio context to prepare the reply and cover letter before WhatsApp delivery.

outcome

  • The pipeline turns daily board scanning into a short approval queue with role-aware outreach ready.
  • Delivery state, retry pressure, source failures, and provider usage all stay visible instead of buried in logs.
  • The human chooses approve, edit, or skip while the system keeps search and drafting repeatable each morning.

stack

Node.jsMongoDBGroqPuppeteerReddit APIHN AlgoliaGitHub GraphQLnode-cron