Mar 31, 2026 · 31m · y-combinator
This Startup Catches Fraud at Scale · Y Combinator
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of Founder Firesides, Y Combinator Managing Partner Jared Friedman interviews Karine Mellata, Co-founder and CEO of Variance, as the company emerges from stealth with a $21 million Series A announcement. Karine discusses how Variance utilizes autonomous AI agents to automate complex fraud, compliance, and risk reviews for major enterprises while reflecting on founder resilience and AI-native engineering culture.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The partners hold 19.3% of the talking time here. How this is scored →
The partners as informed peer 4.6 Guest teaching 3.4 Guest disagreement 0.3 The partners pushing back 0.6
speaking balance: gold is the partners, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 9:57 Rebuttal of simplistic database ingestion premise
Karine directly counters Jared's assumption that data ingestion is straightforward database extraction with 'I wish it was this simple' and explains real-world fragmentation.
Hardest push from the partners ▶ 9:57 Jared challenges the complexity of unstructured dataJared interrupts to probe whether the data ingestion challenge is overstated, asking why they cannot simply query relational databases directly.
Biggest teaching moment ▶ 10:04 Karine educates on scraping human UIs for enterprise dataKarine breaks down why fraud data is locked behind multiple disjointed internal dashboards, forcing them to deploy headless browser agents to capture UI data.
The partners hold their own ▶ 4:39 Jared breaks down GoFundMe disaster fraud dynamicsJared demonstrates sharp domain comprehension by articulating the exact sequence of opportunistic fake fundraisers following high-profile crises.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The partners as informed peer | Guest teaching | Guest disagreement | The partners pushing back | Why |
|---|---|---|---|---|---|---|
| What Variance Does and Stealth Strategy | 4 | 3 | 0 | 0 | Jared sets a collaborative tone celebrating the Series A announcement and clearly illustrates how GoFundMe fraud works. Karine explains why Variance operates in stealth to avoid tipping off bad actors. | |
| Broader Market Use Cases and AI Agent Architecture | 5 | 4 | 0 | 1 | Jared asks about data acquisition methods and connects them to human analyst behavior. Karine outlines the three core pillars of agentic risk systems and accessing the open web. | |
| Technical Challenges: Ingesting Unstructured UI Data | 4 | 6 | 2 | 3 | Jared pushes back on the difficulty of data ingestion by asking if it is just relational DB queries. Karine corrects him by explaining that enterprise data is scattered across legacy systems and hidden behind human-facing UIs. | |
| Shift from Deterministic Rules to Self-Healing AI Agents | 5 | 4 | 0 | 0 | Jared guides the discussion toward agentic evolution beyond legacy rules engines. Karine details how self-healing agent architectures uncovered multi-entity state-sponsored election influence rings. | |
| Real-World Impact, Lean Team Leverage, and Hiring | 4 | 3 | 0 | 0 | Jared and Karine discuss maintaining extreme leverage with a lean team of five engineers. Karine shares unexpected lessons regarding the necessity of human review UIs and non-technical staff using Cursor. | |
| Origin Story: From Apple to YC and Enterprise Sales | 5 | 2 | 0 | 0 | Jared contributes context about IAC's portfolio and highlights the difficulty of landing enterprise customers before ChatGPT existed. Karine explains how their Apple fraud background earned early trust. | |
| Founder Crisis: Facing Adversity and Conviction in Mission | 5 | 2 | 0 | 0 | Jared synthesizes how Variance differed from typical pivoting startups due to domain conviction. Karine recounts surviving a severe accident and finding renewed dedication to their mission. |
Statements from this episode (14)
Friedman: Variance emerges from stealth with $21M Series A
“We are announcing that Variance is coming out of stealth and announcing their twenty-one million dollar Series A, and I am really excited to be here with Karine, the co-founder of Variance.”
Mellata: Variance Reviews GoFundMe Fundraisers and Fortune 50 Seller Verifications
“We've been working with GoFundMe, for instance, to review all of their fundraisers at scale, some Fortune 50 for verifying all of their sellers and complex UBO verifications, and today coming out of stealth.”
Mellata: GoFundMe uses Variance AI agents for fraud and sanctions compliance
“So Variance is used to verify, for instance, if a fundraiser is going to be built for A military operation, for instance, or if you're building a fundraiser for a crisis that you were not part of, which would be fraudulent, well, GoFundMe sort of has the respo…”
Mellata: Automating KYC and KYB Requires Only Three Core Building Blocks
“There's really only three building blocks that you need. So you have the compliance documents, the standard operating procedure, so what the company deems As necessary to verify at the onboarding level or at any other parts of the life cycle of that entity. On…”
Mellata: Variance agents can scrape legacy human review UIs via browser
“So at the moment, the way to integrate with variants can be reverse ETL, API, but very recently we've been awarded this third way, which is spinning up a browser, opening up a really old review tool that was built for a human, pulling that data, and then reaso…”
Mellata: AI Agents Eliminate the Need for Specialized Fraud Classifiers
“You don't need a classifier anymore because AI agents are able to read a set of standard operating procedure and reason over an image or reason over sort of unstructured data and know that this is possibly chargeback fraud. You don't need a specialized classif…”
Mellata: Variance detected state-sponsored election fraud for a Fortune 500 client
“And throughout the elections, because our AI agents had access to the context of entities in relation to other entities, so How does this user fit into all of the other users that we're looking at? We were able to detect really complex fraud rings of especiall…”
Mellata: Variance operates with 12 total employees and five software engineers
“So we're 12 we have five software engineers, so we've been, we've remained very, very lean.”
Mellata: AI agent triage leaves a complex 1% requiring investigative dashboards
“What we found out is that because AI agents are able to take on the simplest part of the workflow, so they're able to triage 99% of cases, that one percent is usually going to be the most complex cases that need to be reviewed by a human, and you need a really…”
Mellata: Variance's five engineers output software equivalent to 25 people
“I would say we're five, but I think in terms of software output, we're probably closer to a 25 people team.”
Mellata: Variance's non-technical CSM ships features autonomously with Cursor
“Our customer success manager, who's entirely non-technical, but interfaces with enterprise customers on a day-to-day basis, now gets to take on feature requests, especially the simple ones, directly give them to Cursor, a Cursor agent, and then directly be abl…”
Mellata: IAC was Variance's first enterprise customer
“Our first customer was a company called IAC. They're a publicly traded company.”
Mellata: OpenAI model upgrades cut Variance's pilot costs tenfold
“As we were running this pilot for this first customer, sort of OpenAI was coming out with new models in the middle of the pilot, which was changing, one, our cost structure by a 10 x factor, but also was changing our performance quite a lot.”