This Startup Secretly Detects Fraud For Fortune 500s
In a Nutshell
Variance emerged from 3 years in stealth with $21M Series A to deliver AI agents automating fraud detection, compliance reviews, and KYB verifications for Fortune 500s like GoFundMe (validating all fundraisers) and marketplaces, using unstructured data, SOPs, and web tools to outperform human processes in cat-and-mouse fraud battles. Founders Karine (data) and Michael (ML), ex-Apple fraud team, built a 12-person AI-maximalist team in SF, achieving rapid enterprise growth (e.g., doubling revenue monthly by July 2024) despite challenges like Karine's severe bike accident that broke her spine and leg, testing the company's resilience. Their unwavering sense of duty to solve fraud at scale with agentic AI—sticking to their Day 0 vision through self-healing systems—drove customer trust and survival.
These notes were generated by AI and may contain inaccuracies.
Variance is coming out of stealth after building in stealth for 3 years, announcing $21 million Series A. Karine is co-founder with Michael.
Variance builds purpose-built AI agents for risk and compliance, automating content review, fraud reviews, identity reviews at scale. Powers Fortune 500s and marketplaces, including GoFundMe (reviews all fundraisers), Fortune 50 for seller verification and complex UBO verifications.
Deals with sensitive data and issues; builds systems used by bad guys but for good guys. Hard to market use cases without creating more fraud/abuse. Customers in cat-and-mouse game with fraudsters; Variance is their secret weapon, operating in the shadows. Will remain somewhat shadowy post-Series A.
Products powered by Variance are used indirectly by everyone, as infrastructure.
GoFundMe: payments platform with strict compliance (liable for payments to bad actors). Variance AI agents verify fundraisers (e.g., not for military operations, crises user not part of, sanctioned countries). Ensures money goes to claimed person.
Typical Fraud on GoFundMe
Crisis-driven spikes post-events/natural disasters. Example: Spike in fundraisers claiming relation to Charlie Kirk's family after his murder; most fraudsters posing as family to siphon donations.
Variance AI agents use behavioral signals, identity info, account history, fundraiser details (images, bio), GoFundMe terms of service to decide allowance. Previously human analysts; now fully automated, consistent. Every GoFundMe signup validated by Variance before live.
- Complex identity reviews for marketplaces/gig platforms (e.g., delivery drivers: selfies, driver's license, per company SOPs).
- KYB verifications: Link person to business (e.g., Karine to Decoy Technologies/Karine Muladah). Builds graphs at scale; detects shell companies, sanctioned nodes, adverse media, money laundering.
Currently manual at scale.
Three blocks: compliance documents, standard operating procedures (SOPs), tools/data (internal/external).
Data split: internal customer data (Variance connects sources, pools unstructured data); external (hundreds of business registries worldwide, open web).
Web access key enabler; mimics human Googling names for fishy signals in abuse graphs.
Core: Data - pull petabytes unstructured data from different sources without schema.
Example (fundraiser): Scattered across user identity, login behaviors, devices, PII, business info, fundraiser history - in 5-10 systems, some behind UIs.
Variance scrapes UIs built for humans via browser agents (reverse ETL, API, or spin up browser on old review tools).
At Apple (prelims): Patchwork of deterministic rules (e.g., transaction >$1,000), classifiers for specific abuse, humans for context.
Problems: Slow/inconsistent human feedback loops; can't achieve self-healing in dynamic fraud environment with adversaries.
AI agents close loop: self-healing/resilient; materialize rules features; reason over SOPs/images/unstructured data (no classifiers needed); replace humans.
Enables shipping faster, new product lines without bottlenecks.
Superhuman Example
Fortune 500 with politically exposed communities: During elections, detected complex fraud rings (state-sponsored actors pushing narratives) via entity context/relations, on-the-fly features, dynamic tool calls. Impossible with isolated classifiers.
Implications: Detects misinformation, threats of physical harm (escalates to law enforcement), prevents violence.
Team of 12; 5 software engineers processing petabytes, fully automating decisions for largest companies. 2x founders. Strong ownership culture; in-person SF daily; product-focused; engineers own problems fully.
Team members expert in areas (e.g., Luke from Meter, top LLM evals).
AI coding maximalists: Engineers manage AI agent teams (3 monitors); output like 25-person team. Review PRs. Non-technical CSM ships simple features via Cursor agents autonomously.
Initially built full end-to-end decisioning; realized AI triages 99%, humans need dashboards for 1% complex cases. Front-end critical.
Karine (data engineer), Michael (ML engineer) met at Apple fraud team (served iMessage, iCloud). Symbiotic: Michael's ML dispatched via Karine's streaming jobs.
Wanted self-healing product; applied to YC.
First Customer
IAC (publicly traded; brands like Care.com, Ask Media Group). Reviewed marketing content for compliance (hard for classifiers/regex; e.g., legal defense advice). Previously large outsourced human team bottlenecking growth.
First to deliver with LLMs. Started pre-ChatGPT; GPT-4 during YC batch/pilot (10x cost drop, performance gains). Took 8 months to land IAC; burning pain point essential for early trust (problem "on fire").
First customers bet on founders' understanding to evolve product.
You really did it the hard way because you went enterprise from the very beginning. Have there been any other hard challenges of building this company?
Starting a company tests you in a lot of different ways. We have a really interesting story. After we got on board a lot of great customers in the trust and safety space—Medium, GoFundMe, Redbubble—around July 2024 was one of the times where the company was growing rapidly. We were onboarding more and more enterprise use cases. During that month, our revenue was doubling within the month and then doubling the month after. It was really exciting, very step function because it's enterprises. We had just wrapped up one of the largest trust and safety conferences called Trustcon in San Francisco. A couple days after, we had worked so hard for this conference—12 hours a day, super tired. I was going back to the office on a Sunday afternoon on the bike lane and a truck hit me. That was a really crazy experience to go through as a founder. The company was doing well, but at the end of the day, we were a 10 people team, and the CEO just gets hit by a truck.
How badly were you injured?
I broke my spine, broke my leg, and was hospitalized for about 10 days. I couldn't walk for about 10 days. For a year and a half as a founder, you're moving so fast, working every day. We never took days off. Even if you do, you're still pretty much on. All of a sudden, you're in a bed and you can't move.
3 or 4 days after, Michael came to visit me in the hospital. Michael was super anxious, one for my health, but also trying to understand what that even meant—the CEO has a bust factor of one. We only have engineers and me running all the sales and customer relationships. It was all founder-led sales; I was the only person doing sales. The rest of the team was only engineers.
Michael brought a Norman Foster book, the architect of Apple Park, as a gift. He was sitting next to my hospital bed holding the book. We were both in silence for a couple of minutes because we didn't know what to say. He laughed and said,
"Well, this is going to make a really good scene in our IPO movie."
I was like, "Yeah, it's a good way to view it." We laughed about it, but I was probably out of commission for about a week, and two or 3 weeks after that. It put a lot of stress on the company. There were a few moments where Michael was wondering if this was the end. He would tell me and repeat the story of Steve Wozniak who went through the plane crash, left Apple, and went back to Berkeley. There was a feeling that maybe this was going to be the end of the company and maybe we just needed to part ways.
There was a really deep feeling that it was not the end. It's an interesting challenge, a hurdle you need to go over, but it just doesn't feel like the end at all. It feels like there's so much more to come. I definitely felt that. Michael felt that. We got to walk again. Now I can walk again. We learned that we definitely need to scale me so that hopefully this doesn't happen again.
One thing that strikes me about your story is the extent to which you guys have had a very strong opinion from the very beginning about what to build and who to build it for. A lot of founders come into YC with some initial idea, but it's just a hypothesis, and then they pivot through many different ideas. New models drop, AI changes, and they might change what they're working on several times to fit whatever the cool thing is. You guys have been the opposite. You came in day zero with a very strong opinion of what to build. You'd seen the problem firsthand, and the whole company has just been that initial hypothesis playing out. Is that perhaps part of why it didn't seem like it could be over yet because you really had to see it through?
Yes. If we go back to why Michael and I started the company, we had a very specific pair of skill sets in fraud and understood what the industry looked like. We had a lot of issues with how these problems were solved at scale. From the beginning, we always felt a really strong sense of duty to put our very specific and quite rare pair of skill sets to the good of the industry.
It was almost like a sense of duty. To us, it was never really about starting a company for any problem sets. We wanted to solve that problem. We knew the technology was going to evolve, and we were so lucky—LLMs got so good. Now we have agentic systems, agent harnesses that are able to fully solve this problem end to end. That strong sense of duty is what kept Michael and I going throughout the years. That's something that resonates deeply with customers. When they meet us, they see these founders really deeply trying to solve the problem they're seeing on a day-to-day basis—once because they've seen it before, but also because it's something that is doable if you put enough care into building the right engineering system in the right ways.
All right, that might be a great note to end on. Thank you so much, Karine.
Thank you.
Keep Y Combinator in your library
Save the videos and channels worth coming back to, and find them again in one place.
