# Aleksei Romanov — extended summary

Hand-written, public, current as of September 2026. This is a one-page biography,
not a knowledge base: it says what he has done at the level of a strong bio and
stops there. Short version: https://romanov.app/llms.txt

Contact: hello@romanov.app · https://linkedin.com/in/romanov-as ·
https://github.com/change101

---

## In one line

AI product leader and hands-on builder who ships conversational AI that completes
real transactions — currently at Ria Money Transfer, co-founder and CTO of Sigtrip.

## In a paragraph

Aleksei Romanov is an AI-native product manager based in Irvine, California. At
Ria Money Transfer (Euronet Worldwide, NASDAQ: EEFT) he owns a net-new
conversational money-transfer channel that runs inside WhatsApp — LLM-driven
interaction wrapped around deterministic payments, identity, fraud, and compliance
flows. He is also co-founder and CTO of Sigtrip, which built the first
OpenAI-approved hotel booking experience to complete reservations directly inside
ChatGPT. Before that he co-founded two more companies. He writes production code
and submits PRs alongside his engineering teams.

## The throughline

His specialty is turning AI conversations into completed transactions rather than
answers. Most AI product work stops at generating text; the hard part — identity,
money movement, compliance, deterministic handoff, what happens when the model is
wrong — starts exactly where the demo ends.

He has now built that twice, in two regulated domains. At Ria, inside a
thirty-year-old global financial services company, he took a conversational
money-transfer channel from a blank page to the company's fastest-growing new
channel. At Sigtrip he built an MCP-native transaction layer that lets a traveler
search live hotel availability, pick a room, and confirm a booking without leaving
an AI conversation, with the reservation written into the hotel's own system of
record.

What makes the profile unusual is the combination rather than any single part: a
PM who owns strategy and roadmap for multiple teams, and who also ships production
Python, designs retrieval and evaluation pipelines, and builds his own internal
tooling.

---

## Ria Money Transfer — Product Manager, Conversational AI & Payments
**Irvine, CA · September 2024 – present.** Euronet Worldwide (NASDAQ: EEFT).
Title of record is Product Manager; "Conversational AI & Payments" describes the
work.

**The WhatsApp money-transfer channel.** Ria's net-new channel for sending
cross-border transfers inside WhatsApp. The product problem is that a money
transfer is not a chat — it is a compliance-gated, identity-verified,
fraud-screened financial transaction that has to be exactly right, delivered
through an interface where people type in plain language. The architecture
reflects that: an LLM drives the conversation, and everything touching money,
identity, or compliance runs through deterministic flows the model hands off to
rather than decides.

He started from a blank page: product vision, LLM conversation design under KYC
and PII constraints, PRDs written with ML and backend engineers, WhatsApp Business
API compliance, and simultaneous cross-border go-to-market for two sending
markets. The channel launched in early 2026, reached 100,000 users in its first
half-year, and is currently compounding 50–100% month over month on that base. It
is live for sending in the US and South Africa, with transfers reaching more than
190 recipient countries. Notable that this happened inside a thirty-year-old
public financial services company: a net-new digital distribution channel, built
at startup pace, under public-company compliance constraints.

**Replacing a vendor with an in-house build.** A third-party platform in the
channel's critical path was expensive and the source of recurring performance
incidents; at projected scale the per-message cost would have made the channel's
unit economics unviable. He ran the make-versus-buy analysis, modeled the cost
trajectory across three growth scenarios, presented the break-even case to
leadership, and drove delivery of a custom replacement — accepting a roughly
three-times-longer timeline and negotiating a phased migration to de-risk it. The
result was a 90% infrastructure cost reduction, savings in the seven figures
annually, and the root cause of the production incidents eliminated.

**An ML-driven onboarding redesign.** Owned while running the Acquisition team,
on Ria's core product rather than the WhatsApp channel. Standard funnel analytics
showed where users dropped but not why, so he applied NLP models to thousands of
customer-feedback entries and surfaced friction patterns manual review had missed.
The analysis pointed at the identity-verification step rather than the form fields
everyone had assumed. He argued for redesigning that single step instead of the
whole funnel, scoped it as a two-week sprint, and measured over a sixty-day
cohort: retention up 16%, activation up 12%.

**Scope.** Led product strategy and roadmap across two product teams — Acquisition
and AI Growth Channels — comprising eighteen engineers in frontend, backend, ML
and data. Cross-functional scope rather than direct reports. He also defined the
responsible-AI guardrails for an LLM driving financial transactions: what the
model may do, what data it can reach, PII and data-access boundaries, and where
control must hand off deterministically for compliance-sensitive actions.

## Sigtrip — Co-founder & CTO
**Remote · June 2023 – present.** AI-native hotel commerce infrastructure.

Sigtrip's thesis is that AI assistants are the third major distribution reset in
travel, after the GDS and the OTAs, and that hotels have no infrastructure for it.
It is a three-sided marketplace: hotels, travelers, and AI assistants as the new
point of discovery. It began as a consumer AI trip-planning iOS app and pivoted to
B2B infrastructure; the pivot is the story, and the consumer app is discontinued.

**AI Direct Bookings.** An MCP-native transaction layer that abstracts fragmented
hospitality infrastructure into standardized agent tools — property knowledge,
inventory, rates, booking, confirmation, cancellation. A traveler can search live
availability, select a room, and confirm without leaving the AI conversation, with
the reservation written into the hotel's own system of record so the guest
relationship stays with the hotel. Built on this layer: the first OpenAI-approved
hotel booking experience to complete reservations directly inside ChatGPT.
Bookings carry no OTA commission; Sigtrip charges its own flat per-booking fee
instead of a cut of the rate.

**AI Visibility.** A pipeline that queries every major AI engine daily with real
traveler prompts to measure how ChatGPT, Perplexity, Google AI and Claude describe
and recommend a property, tied through to confirmed bookings. Aleksei designed the
scoring model end to end — a visibility index, mention rate, share of voice and
booking-path integrity — over persona-simulated prompts, using two-pass detection:
literal name matching plus an LLM reading for paraphrase, rank and competitive
context.

**Published research.** An open census of which hotels ChatGPT names when
travelers ask where to stay, live across the top US hotel markets, with published
methodology, sample sizes and confidence ranges, per-market data under CC BY 4.0,
a conflict-of-interest disclosure and a correction process. No paid placement — a
deliberate credibility constraint on a metric the company also sells against.

What Sigtrip is not: not an OTA or a booking site — it does not sell rooms or hold
inventory; and not affiliated with OpenAI, Anthropic, Google or Perplexity, whose
public answers it measures.

## Stuva — Co-founder & Head of Product
**San Francisco, CA · March 2022 – September 2024.** AI-powered real estate
marketplace, wound down in 2024 after running out of funding — a completed 0→1
that reached real usage, not an ongoing company. Full product lifecycle from idea
to launch in eight months. Designed a swipe-to-match experience and built the
recommendation engine on reinforcement learning, collaborative filtering and
content-based ranking, learning from swipe patterns, dwell time and explicit
saves. Also applied ML to property valuation and listing classification, and built
performance-critical infrastructure including a 60fps media feed over thousands of
image-heavy listings.

## WestLA — Co-founder & Head of Product
**Los Angeles, CA · March 2020 – March 2022.** Cross-border e-commerce platform
connecting US and EU retailers with customers across CIS markets. Owned product
strategy and built the underlying payments, order and international logistics
workflows from concept through production.

## Earlier
**BIC Group — Management Consultant, 2019.** Statistical demand and pricing models
for European steel companies. **Aton Investment Group — Sales & Trading, 2017 –
2018.** An independent investment bank. The quantitative and unit-economics
instincts that show up in his product work start here.

---

## Skills, with where they were demonstrated

**AI and ML.** LLM product development (Ria channel, Sigtrip). Agentic AI, tool
use and MCP (Sigtrip's transaction layer). LLM evaluation — a multi-engine harness
with persona simulation, paraphrase-tolerant two-pass detection, and statistical
treatment of engine non-determinism. Retrieval pipelines for grounded property and
policy answers. AI governance and guardrails — PII and data-access boundaries for
an LLM driving financial transactions. Recommendation systems and reinforcement
learning (Stuva). Applied ML and NLP.

**Product.** 0→1 development. Cross-functional leadership across eighteen
engineers and two teams. Build-versus-buy analysis with a three-scenario cost
model. Technical specification written with ML and backend engineers.
Go-to-market for a simultaneous two-market launch. Growth, retention and
activation. Marketplace and regulated-product work.

**Engineering.** Python, SQL and PostgreSQL, SwiftUI and iOS, AWS, real-time and
event-driven systems, Supabase and Firebase, analytics instrumentation. Ships
production code and submits PRs. Public contribution activity is visible at
https://github.com/change101 and rendered as a live heatmap on romanov.app.

**Domains.** FinTech and payments, consumer marketplaces, AI-first products,
travel and hospitality, international and multi-market, trust and safety. Open
across industries wherever AI has to act rather than only answer; the one he is not
pursuing is healthcare and life sciences.

## Practical

Based in Irvine, California. Holds an EAD and requires no visa sponsorship, now or
in the future. Open to remote and hybrid, and to the right on-site or relocation
situation. Looking for full-time senior or staff product roles that are AI-first
and genuinely technical, and a strong fit for product-engineering hybrid and
founding-PM roles. What he optimizes for: a surface he owns end to end, a team
technical enough to engage with him at the level of the architecture, and problems
where AI completes real actions with consequences. He is hands-on by default and
writes production code, which is unusual for a PM and is evidence of that depth
rather than something a role has to offer him.

Reach him at hello@romanov.app.