About neonVestneonVest is an AI-powered fundraising and investor intelligence platform. We help founders raise capital through a curated network of 720+ active investors — our Superchargers, who take 1:1 meetings with companies we introduce — and a matching layer that puts a company in front of the investors who will actually care about it rather than blasting everyone and hoping.

We're headquartered in New York, most of our clients are US-based, and the team is small enough that everyone owns something real.

The shape of itThis starts as a defined three-month build — the investor matching service, end to end — and is intended to continue. We're being explicit about that rather than vague: we know what we want built first, we'd rather agree on it clearly than hand you an open-ended remit, and if it goes well the platform work after it is a year of runway at least.

What continuation depends on is stated below, so you can judge it for yourself rather than hoping.

The roleWe're rebuilding the core of our product — investor matching — as an AI-native service, and you'd be the person building it.

Today, matching is a filtered database query. We want a system that reads a company, understands what an investor actually looks for, ranks the fit, and explains its reasoning. That's the first thing you'd build, and it's the piece that matters most commercially.

The architecture will be specified before you start. Data model, service boundaries, how retrieval and reasoning fit together, how we evaluate it. You'd be building against a design rather than inventing one — and you'd be expected to push back on it where it's wrong.

What you'd build, in orderThe investor data layer. Our investor data currently sits in two places: structured attributes in Airtable — sector, stage, cheque size, geography — and free-text descriptions plus years of match history in a MySQL database. First job is merging them into one clean, keyed dataset. Unglamorous, and everything else depends on it.

The matching service.



Embeddings over investor descriptions and company materials, retrieval to narrow the field, structured filters for hard constraints, and an LLM layer that reasons about fit and produces an explanation. Exposed as an API our existing platform calls.

The evaluation harness. We have hundreds of past matches with known outcomes — meetings taken, feedback given, conversions. That's your ground truth. Every change to matching gets scored against it. "It feels better" is not a result.

Then, incrementally, more of the platform. Once matching is live, we move the next capability across, and the next. Our existing PHP application keeps running the business throughout and gets switched off only when nothing calls it any more.

And keeping the old system alive meanwhile. A CodeIgniter application from around 2021. Not glamorous, genuinely necessary — it carries real client and investor data.

What we're looking for5–8 years building production software. You've shipped things people depend on and maintained them afterwards.

Strong Python or TypeScript. The new service will be one of those.

You've worked with LLMs and retrieval in production, not just demos. You know why RAG pipelines disappoint in practice and what to do about it. Research-level ML isn't needed; applied judgment is.

You build with AI agents and know where they lie. We use them heavily and expect you to be much faster because of it. The skill we're actually hiring for is catching the confident, plausible, subtly wrong output before it ships.

Data engineering instincts. Our investor data has duplicate tables, contacts crammed into free-text fields,



and years of drift. Making it clean and queryable is the job, not a preamble to it.

You can work in inherited PHP without wanting to rewrite it. You don't have to enjoy CodeIgniter. You do have to keep it running and resist touching what doesn't need touching.

You can write. Small distributed team, mostly async. Explaining a technical trade-off to a non-technical founder in three sentences is part of the role.

Nice to have

- Vector databases, embedding models, evaluation frameworks
- Experience replacing a legacy system piece by piece rather than in one jump
- AWS — the platform runs on EC2
- Venture, fundraising or financial services exposure

What you'd be walking into, honestlyThe existing application is about five years old and predates everyone currently here. There's no version control, no test suite, and until recently no copy of the code outside the production server. The operating system is out of support. We're saying so because it's the job. If that reads as a mess to avoid, we're the wrong place. If it reads as a transparent runway — real data, real users, real revenue, and no incumbent architecture to argue with — it's the best kind of first ninety days.

Practical

- Full-time, remote
- Meaningful overlap with US Eastern hours
- ₹17–23 LPA, paid monthly in USD ($1,500–2,000/month), with room to grow as the platform doesInitial engagement of three months, intended to continue.
- What that depends on: the matching service is live and measurably better than what it replaces, and you've been straightforward about what's working and what isn't. Not on hours logged or lines shipped.
- You'd work directly with the executive team rather than three layers below them

How to applySend your CV, and a short answer to this: Tell us about a time an AI coding assistant produced something that looked right and wasn't. How did you catch it, and what do you do differently now?

📌 Founding Engineer — AI Platform | neonVest (India)
🏢 NeonVest
📍 India