Challenge
Inbound pitch decks arrived by email and cloud links faster than analysts could triage them. Manual copy-paste into trackers created duplicates, blind spots and burnout.
Approach
We treated deal intake as a production pipeline: schedule email watchers, parse every attachment format that shows up, enrich with web intelligence, then emit structured records with clear failure modes.
Solution
A NestJS platform with GPT-4 document understanding, Google Workspace integrations, DocSend handling, founder/market research hooks and APIs/webhooks for the rest of the firm’s stack.
Technology
TypeScript, NestJS, OpenAI GPT-4, PostgreSQL, Docker, Puppeteer/Cheerio, SERP/Apify, OAuth 2.0 — modular services with logging you can audit.
Outcome
~90% of repetitive intake steps automated; pitch review ~10× faster than manual processing — so partners spend time on judgment, not inbox archaeology.
Similar problem?
