Black Leo Ventures
A platform that scores startups against 6,000+ investors and incubators, then runs and tracks their email outreach campaigns end to end.
git clone git@github.com:the-client/black-leo-ventures.git
Cloning into 'black-leo-ventures'...
fatal: could not read from remote repository.
It's on the client's Vercel, not yours.
hint: the code is private, the product isn't. It's running live at investor-outreach-platform-roan.vercel.app ↗
- Role
- Lead Full-Stack Engineer
- Tools
- Next.js · TypeScript · Firebase
- Type
- Investor Outreach
- Year
- 2025 to 2026

01 · Overview
Matching, campaigns and delivery tracking
I took over this project after the client rejected the first build, and led it as full-stack engineer, working directly with the client and my company's CEO. The job was to make automated investor outreach actually work: match startups against 6,000+ investors and incubators, schedule campaigns, send through each client's own SMTP account and report exactly who was matched, reached, opened and replied.
I owned the matching algorithm, the campaign and cron pipeline, credential encryption and reporting, and directed one engineer who built the investor-facing deal room.

02 · Problem statement
Manual outreach doesn't scale
Outreach used to be manual: pick investors by hand from a spreadsheet, write each email, send it and guess what happened next. Nothing matched a startup's sector, stage, location and ticket size to the right investors, and nothing tracked opens or replies.
The goal was to remove every manual step: a startup submits its profile and pitch deck, gets scored against the database, runs a scheduled campaign nobody has to babysit, and gets a report of who was reached and how they responded.
- Manual matching
- No reply tracking
- Spreadsheet workflow
03 · Architecture
A serverless monolith on Vercel
The whole system is one Next.js app: the same codebase serves the admin dashboard and the API routes, and Vercel deploys each route as its own serverless function. Firestore is the only database, Firebase Auth covers admins, sub-admins, clients and investors, and every route checks the caller's token and role first.
Mail goes out through Nodemailer on each client's own SMTP account, a tracking pixel records opens, and an IMAP job reads the client's inbox to catch replies. Gemini rewrites campaign copy and scores pitch decks. Scheduled jobs don't run on Vercel at all: GitHub Actions calls the send, reply-check and stats endpoints on their own schedules, authenticated with a shared secret.
04 · Build
220,000 reads a day down to 5,000
The hardest problem wasn't matching, it was tracking every recipient's state correctly while keeping it cheap. Firestore bills per read, and the app was doing about 220,000 reads a day for just two campaigns: the dashboard recounted every recipient on each load, one campaign page cost about 6,000 reads to open, and the reply checker ran far more often than needed.
Reply matching has its own edge cases: direct replies, forwards answered by a colleague, repeat replies and auto-replies all have to update the right record. I replaced recounts with real-time counters, added ETag caching so an unchanged refresh costs zero reads, checked exact email matches first since most replies are direct, and made the cron skip campaigns that were just sent or already done. Reads fell to about 5,000 a day for the same two campaigns, with no feature removed.
- Real-time counters
- ETag caching
- Lifecycle-aware cron

05 · Project results
97% fewer database reads
Daily Firestore reads dropped from about 220,000 to about 5,000 for the same two campaigns without losing any tracking. The platform is live, matching startups against 6,000+ investors and incubators, and has sent 6,900+ outreach emails for its clients.
- Fewer database reads
- 97%
- Outreach emails sent
- 6,900+
- Investors and incubators
- 6,000+


What shipped
- Took over after the client rejected the first build, and rebuilt it as lead engineer.
- Built a weighted matching engine that scores startups against 6,000+ investors and incubators on sector, stage, location and ticket size.
- Automated campaign delivery through each client's own SMTP account, with per-email open tracking and IMAP reply detection.
- Cut daily Firestore reads from about 220,000 to about 5,000 with real-time counters, HTTP caching and lifecycle-aware cron scheduling.
- Integrated Google Gemini to rewrite campaign copy and score pitch decks for investment readiness.
- Directed one engineer's delivery of the investor-facing deal room.
