Case study 01 / 06

FundReachr

Fundraising · SaaS

An AI investor-outreach SaaS, built and shipped end to end

Data engineering
Software engineering
Cloud & infrastructure

fundreachr.com

platform.fundreachr.comDemo data

Key figures

Investor data synced into the app: Daily
Canonical record per investor: 1
Sectors in the investor taxonomy: 43
Languages for AI-written messages: 11

The brief

The challenge
Take a fundraising-automation idea to a live product with no codebase, no team and no infrastructure, for a promise that depends on an investor database that is curated, enriched and always fresh.
What we built
The whole platform: an investor-data engine that collects, classifies, enriches and de-duplicates investors into BigQuery; a Flask API with Auth0, Stripe credit billing and AI message generation; email-delivery workers; four React apps and a browser extension, running on Google Cloud with Postgres on Neon.
Outcome
A live SaaS on Google Cloud whose billing, authentication and outreach were re-engineered as the product found its model.

Expertise

What it took.

Data engineering
- An event-driven investor-data pipeline on Cloud Tasks: collect → LLM classification → multi-provider enrichment → taxonomy mapping, landing in a layered BigQuery warehouse.
- Identity resolution that gives each investor one canonical ID whichever source found them, and caches (30-day search, 90-day enrichment, permanent LLM answers) so paid lookups and LLM calls are not repeated within their window.
- A daily BigQuery-to-Postgres sync behind investor search built on JSONB and JSONPath, where free-text cheque sizes are parsed inside SQL.
Software engineering
- A unit-tested Flask API with Auth0 sign-in behind a role-and-ownership authorization layer.
- Stripe credit packs with idempotent webhooks, a wallet and an auditable credit ledger; discount codes mirrored as Stripe coupons.
- Four React 19 apps (customer platform, admin back office, internal CRM, marketing site) and a Manifest V3 extension that fills investors' web forms.
Cloud & infrastructure
- Cloud Run services and jobs, Cloud Functions, Cloud Tasks and Cloud Scheduler, with OIDC tokens on the scheduled and queued calls into the private, IAM-only workers.
- Cloud Build pipelines for every deployable service (Artifact Registry, Secret Manager), with dev → staging → main promotion on the core API, customer app and outreach workers.
- Queue-based worker pools that claim work with SKIP LOCKED and recover from crashes, so campaigns of any size never block the API.

Project 01

Customer platform

React 19 · Flask API · Cloud Run

Where founders define an audience, see the reach and credit cost before anything is generated, review every AI-written message per channel, send from their own Brevo account and follow delivery, opens and clicks.

Goal: Package FundReachr's investor database and AI writing as a self-serve, pay-per-use product.
Highlights: Three-step campaign wizard with autosaved drafts and live reach and cost estimates over the investor database
Per-message control: edit or regenerate any message, and send emails one by one or in bulk
Stripe credit packs, discount codes, receipts and a credit ledger
Analytics driven by Brevo webhooks through each message's lifecycle
Stack: React 19
Flask
PostgreSQL
Auth0
Stripe
Brevo

AI-written investor email in the FundReachr campaign workspace — Every investor gets a tailored email that the founder can preview, edit, regenerate or send, with delivery, open and click tracking.

Campaign launch review with estimated reach and credit cost — Launch review: reach and credit cost estimated against the investor database before a single message is generated.

FundReachr analytics with credit and email KPIs and a credit balance chart — Founder analytics built on the credit ledger and email engagement events.

Campaign wizard audience step with investor filters — Audience targeting that mirrors the investor data model: investor type, stage, ticket size and industry.

Grid of fundraising campaigns with progress bars — Every campaign's lifecycle at a glance, from draft and AI generation to sending and completion.

Project 02

Investor-data engine

Python · Cloud Run · Cloud Tasks · BigQuery

Builds and refreshes the database founders search. Source adapters collect investor and fund profiles from web search, investor lists, public registries, directories, VC websites and file drops; an LLM screens web-search results and file drops so only real investors are kept, and extracts their thesis; contact details are enriched; duplicates merge into one canonical record; everything maps to a fixed taxonomy.

Goal: Keep a curated, de-duplicated investor database fresh without manual list-building, with paid lookups under control.
Highlights: Event chain on Cloud Tasks: collect → classify → enrich → map
43 sectors, 9 funding stages and structured geographies
Caching tiers for search, enrichment and LLM answers
A Playwright crawler that collected pitch forms from VC websites and mapped their fields to the founder's data with an LLM
Stack: Python
Flask
BigQuery
Datastore
Cloud Tasks
OpenAI
Gemini
Playwright

Diagram of the FundReachr investor-data pipeline — The investor-data engine: scheduled collection, LLM classification, enrichment and identity resolution into BigQuery, synced daily into the app.

Project 03

Outreach engine

Cloud Functions · Cloud Tasks · OpenAI · Brevo

Writes every campaign message in parallel and delivers the emails. Generator workers ask OpenAI for an email or LinkedIn note tailored to each investor's thesis and the founder's pitch; sender workers deliver emails through the founder's own Brevo account, with credit checks and an optional pitch-deck attachment.

Goal: Personalised outreach at scale that survives crashes, pauses when credits run out and never lets two workers send the same message.
Highlights: Postgres as a work queue: SKIP LOCKED claims, stale-work recovery, row-locked completion
Per-platform, per-language choice between AI generation and templates
A browser extension that fills investors' web forms with the AI-written values
Stack: Python
Cloud Functions
Cloud Tasks
OpenAI
Brevo
PostgreSQL

Diagram of the FundReachr outreach engine — How a campaign flows from the founder's launch to generated, sent and tracked messages.

Browser extension auto-filling a VC pitch form — The Form Filler extension completing a fictional fund's pitch form with the AI-written values, debug overlay on.

Project 04

Back office & internal CRM

React 19 · Flask · Cloud Run Jobs

An admin suite to run the SaaS without touching the database (users, credits, packs, Stripe-backed discount codes and review queues for user-proposed corrections to investor records), plus the CRM FundReachr uses to sell itself, sending FR/EN sequences in quota-sized daily batches.

Goal: Operate, support and grow the product with in-house tools instead of manual database work.
Highlights: Field-by-field review of crowd-sourced corrections before they reach the investor database
Credit grants and deductions written to the ledger
Cloud Tasks → Cloud Run Job batch sending within the email provider's daily quota
In-app documentation with Mermaid diagrams of the campaign lifecycle and data model
Stack: React 19
Flask
Alembic
Cloud Run Jobs
Cloud Tasks
Mermaid

Internal CRM campaign page with scheduled sequences — Internal CRM: a sequence scheduled for automated daily sending, every step tracked per founder.

Admin review of proposed changes to an investor record — Admin back office: a user-proposed correction to an investor record, reviewed field by field before it is applied.

Admin dashboard with KPI cards and status breakdowns — Operator dashboard: users, campaigns and pending data reviews at a glance.

Internal CRM prospect pipeline table — Prospect pipeline with stage counts, sequence progress, opens and appointments.

Admin user detail page with credit wallet — Customer account management: access, credit grants written to the ledger, and the active pack.

Internal CRM help page with an entity-relationship diagram and a status chart — Documentation shipped inside the CRM: the data model as a Mermaid diagram, above a live chart of messages by status.

Project 05

Marketing site

React · three.js · Cloud Run

fundreachr.com: a WebGL hero, a three.js visual of the investor network and pack prices read live from the product catalog, plus the scripted Playwright pipeline that records the product demo video.

Goal: Turn founders into sign-ups, with real product footage and pack prices read from the live catalog.
Highlights: Custom three.js network with shader-driven connections
Pricing rendered from the backend catalog, with a resilient fallback
SEO meta on every page, with JSON-LD structured data on the home and FAQ pages
Stack: React
three.js
Tailwind CSS
Cloud Run
Playwright