About Dalphe

We built Dalphe because compliance was the thing nobody at our fund had time for.

Founded in 2024 by a former IR lead and a former regulator. Built for the firms that take their auditor's questions seriously.

2024–26Off-the-shelf AI is built for general helpfulness. Investor relations is the opposite problem: most of the time, the right answer is to refuse, redirect, or require verification. We built a chat assistant whose first instinct is to not answer — and whose second instinct is to write down everything it did.

01 · The wagerThat a private fund manager would rather pay for the constraints than retrofit them later under SEC examination.
02 · The betThat compliance is a database problem more than a model problem — and the database is where the controls belong.
03 · The promiseThat you can ship a chat assistant on your IR page tomorrow without inviting your General Counsel into the loop every week.

In 2023, Maya spent the better part of a year as the head of investor relations for a mid-market private credit fund. Her job, on paper, was to communicate with limited partners. Her job, in practice, was to be a human firewall — re-reading every outbound message, rewriting half of them, and apologizing for the rest.

The thing nobody told her in the offer letter was how much of investor relations is regulatory rather than relational. Performance figures had to be carefully framed. Forward-looking statements needed boilerplate disclaimers. Anything that touched accredited-investor status required verification. The PPM was a 134-page document she was supposed to know by heart and quote from never.

And then, in late 2023, an analyst on the team pasted a passage from the PPM into a ChatGPT window to draft a response. The compliance officer saw the audit log the next morning. The fund settled with the regulator quietly. Nobody got fired but everyone learned a lesson, which was: the model is not the problem. The lack of constraint around the model is.

The naive thing didn't work.

Maya's first instinct, like everyone's, was to write a long policy document. Train the IR team harder. Add a checklist. The checklist didn't survive contact with a slow day and a chatty LP.

Her second instinct, like everyone's, was to ban the chatbot. That worked for about three weeks. Then someone needed to draft a follow-up at 11pm on a Thursday and the rules quietly relaxed.

“The answer wasn't a better prompt. It was a different shape of system — one where the model couldn't do the unsafe thing even if it wanted to. Where the constraints lived in the database, not in the team's discipline.”Maya Okonkwo, co-founder & CEO

She emailed Daniel — a friend from law school who had spent five years at the SEC's Division of Enforcement. They spent six weekends sketching a different design: a chat surface where classification ran before retrieval, where retrieval was filtered at the SQL layer rather than in application code, where the audit log was an append-only table with a database trigger refusing to mutate it. The model was just the last step.

What we built first.

The first version of Dalphe was a Python script and a Postgres database. It classified messages with Claude Haiku. It refused anything it wasn't sure about. It wrote everything down. Maya's old fund was the first customer. The compliance officer said, in writing, that he'd never seen an AI tool he was willing to defend to an examiner. He was willing to defend this one.

We've kept the shape since then. The pipeline still runs in the same order — classify, gate, retrieve, compose, audit — because that order is the product. The audit log is still the part our customers love most, which we did not predict.

Where we are now.

Twenty-three private fund managers run their IR channel on Dalphe today, representing about $14 billion in combined AUM. None have been fined. Several have been examined. All passed with the audit log as their first exhibit.

We're hiring for engineers who care about correctness more than velocity, and a founding solutions person who has sat across the table from a SEC examiner and can explain what makes them comfortable. If that's you, the email at the bottom of this page is real.

What we believe

Six principles that survived the second year.

These are not aspirational. Each one has been load-bearing in a product decision we've actually made — and at least one we've reversed.

i

Fail closed, always.

If we don't know what category a message is, we refuse. If we don't know if a partner integration is healthy, we refuse. If we don't know whether a user is verified, we refuse. Silence is a feature.

Source · pipeline step 1 & 2
ii

Constraints belong in the database.

Anything enforced in application code can be disabled by a hotfix. Anything enforced by a database trigger requires a migration that someone has to write down. Migrations are how compliance survives a sprint.

Source · postgres/migrations/0007
iii

The auditor is a stakeholder.

We design the audit log first and the chat interface second. If a row can't be explained to a regulator, we don't write it. If a row has to be mutable, it's not a row — it's a state machine somewhere else.

Source · weekly auditor design review
iv

Refusal is a product surface.

"I can't help with that" is a hostile experience. "I can't help — here's why, here's what I can do, here's how to get verified" is a product. We treat refusal templates as carefully as we treat onboarding.

Source · refusal template registry
v

Generic isn't compliant.

A horizontal AI tool retrofitted with policies is not the same product as a vertical tool designed around constraints. We will be smaller than horizontal AI companies. That's the deal.

Source · founding memo, March 2024
vi

Boring beats novel.

Our customers are sophisticated. They are unimpressed by a clever demo and very impressed by a Postgres trigger that's been running for 600 days without a single allowed UPDATE. We optimize for the second thing.

Source · 0 mutate events to date
The team

Nine people. Two of whom have been deposed.

We hire engineers who think like compliance officers, and compliance people who think like engineers. The overlap is small. That's why we're nine.

M

Maya Okonkwo

Co-founder & CEO

Former Head of IR at Roundwall Credit Partners ($2.3B AUM). Before that, three years at Bridgewater on the LP-facing team. Wrote the original Dalphe classifier on a Sunday in February 2024 and hasn't slept properly since.

PrevRoundwall Credit · Bridgewater · Wharton MBA
D

Daniel Reyes

Co-founder & CTO

Five years at the SEC, two of them in the Asset Management Unit working enforcement against private fund advisers. Before that, a securities-litigation associate at Sullivan & Cromwell. Designed the audit log because he's seen what bad ones look like.

PrevSEC · Sullivan & Cromwell · Yale Law
A

Aditi Sharma

Founding engineer

Owns the classification and retrieval pipeline. Previously on Stripe's ML platform team. Will not let a feature ship without a benchmark.

J

Jonas Lindqvist

Founding engineer

Postgres maximalist. Wrote the WORM trigger. Previously infrastructure at Linear and a brief, fondly-remembered stint at Citus Data.

P

Priya Walsh

Head of compliance

Eight years as Chief Compliance Officer at two mid-market hedge funds. Knows what an SEC deficiency letter feels like and how to write the response.

T

Tomas Becker

Solutions engineering

Onboards every new fund personally. Previously implementations at Tessell. Holds the record for fastest go-live: 4 hours, 41 minutes.

Where we are

Two offices and a respectful amount of remote.

New York · HQ

40 Worth Street, Suite 802 Tribeca, New York, NY 10013

EST · UTC-5 · 5 of us

Lisbon

Rua da Misericórdia 14, 2º 1200-269 Lisboa, Portugal

WET · UTC+0 · 2 of us

Remote

US-based only, for now. Quarterly all-hands in NYC.

EST overlap req'd · 2 of us

Run your IR channel on infrastructure your auditor will recognize.

Start with the 30-day trial. If you'd rather hear it directly from us, we keep half a day a week for founder calls — no SDR, no qualification form.

Are we hiring? Yes — engineering & founding SE.
Press & analyst?
press@dalphe.com · we respond in 24 hours.
Security disclosure?
security@dalphe.com