Seven YC rejections. $93M raised at a company I helped build. Here's what I learned — and why the best thing that happened to us was failing to build our own product.
The Early Days
Before Proxie, there was Vance — a YC W22 fintech startup that raised $93M. I was part of the engineering team that built the core infrastructure. Before that, I shipped systems serving 50M+ users at Zeta and CRED. The résumé looked good. The itch to build something of my own was unbearable.
So I started building. Again and again and again.
The Rejection Arc
Attempt 1: Brix AI
An autonomous coding agent. The idea was compelling: point AI at a codebase, describe what you want, and it builds it. We got it working — sort of. The code compiled. It even passed tests sometimes. But the gap between 'compiles' and 'production-ready' turned out to be a chasm. YC said no.
Attempt 2: Fintt AI
An India-native AI finance coach. We built a solid prototype. The technology worked. The unit economics didn't. Customer acquisition cost for a personal finance app in India is brutal, and the willingness to pay for financial advice is... let's say developing. YC said no again.
The Pattern
After 7 rejections across multiple ideas, I finally saw the pattern: we were building products and then looking for customers. Every founder makes this mistake. We just made it seven times.
The Pivot Moment
The pivot came from an unlikely source: customer discovery. We finally did what every startup book tells you to do first — we talked to people.
We interviewed 20+ founders and operators. Every conversation revealed the same pain: they needed consulting-grade work but couldn't afford consulting-grade prices. They needed research, analysis, strategy — but $200K for a McKinsey engagement wasn't in the budget.
And here's the thing we noticed: 70% of what they were paying consultants for was work that AI could do better, faster, and cheaper. Research. Analysis. First drafts. Data synthesis. The remaining 30% — strategy, judgment, relationships — needed humans.
Why Proxie Makes Sense
Every failed product taught us something we use daily at Proxie:
- ▸Brix AI taught us that AI-generated output needs human review. Always. This became our quality pipeline.
- ▸Fintt AI taught us that the technology doesn't matter if the business model doesn't work. Proxie's pricing is validated by real client willingness to pay.
- ▸All 7 rejections taught us that customer-first beats product-first. We now start every engagement by understanding the problem, not by building the solution.
- ▸Our engineering background (Vance, Zeta, CRED) gave us the production chops to ship AI systems that actually work at scale.
What's Different Now
Three things changed:
- ▸Customer-first approach: We don't build and then sell. We listen, scope, and then build exactly what's needed.
- ▸Proving value before scaling: We're not chasing growth metrics. We're chasing client outcomes. One great case study is worth more than 1,000 signups.
- ▸Building an agency that ships: Not a product that might find market fit. An agency that delivers consulting-grade work using AI, today, for real clients.
We're a new agency. We have a small but growing client base. But our team has shipped AI systems reaching 50M+ users at companies like Zeta, CRED, and Vance. We've managed ₹250Cr+ in GMV (Gross Merchandise Value) through our code. We're not new to building — we're new to building for others. And it turns out, that's what we should have been doing all along.
Be Our Next Success Story
Seven rejections taught us that the best technology in the world doesn't matter without a customer who needs it. If you're that customer — if you need consulting-grade work at software prices — we'd love to talk. Not to sell you something, but to understand your problem first. That's the lesson that took 7 tries to learn.