Peter Piatetsky is the CEO and co-founder of Castellum.AI, which started in 2019 as a primary source for sanctions, PEP and adverse media data and now builds AI agents that resolve compliance alerts for banks and fintechs. A former intelligence analyst, OFAC analyst and bank compliance head, he has seen the alert overload problem from every side. We talk about what actually happens when an alert fires, why the real question today is build and maintain versus buy, and what Castellum's benchmark shows about LLMs with and without a harness.

What We Covered

• From intelligence analyst in Afghanistan to OFAC
• Becoming lead Russia analyst when Russia first invaded Ukraine
• Running US compliance for a major Asian bank
• Why Castellum chose to be a primary data source
• Going modular to meet clients where they are
• What happens when an alert hits the Arbiter agents
• Training an agent on a client's policies and procedures in 10 business days
• Why the AI reliability debate is over
• Build and maintain versus buy
• Benchmarking nine models with and without a harness
• Running inside LexisNexis Bridger and Fircosoft
• How criminals are using AI at scale

Key Takeaways

• The harness matters more than the model. In Castellum's testing of nine models, top LLMs on their own were accurate about 70% of the time when asked to read policies, screen, create and adjudicate an alert. With Castellum's harness of specialized data, training, regression testing and QA/QC, accuracy rose to 99% and consistency got close to 100%.
• The question is no longer whether AI can do this. Within a year, buyers have moved from doubting AI agents to asking whether they have budget this year or next. The common objection now is "why not just use Claude," which Piatetsky frames as build and maintain versus buy, including retesting every new model on hundreds or thousands of cases.
• Sit on top of what banks already have. Castellum's agents run inside existing screening systems such as LexisNexis Bridger and Fircosoft, so clients avoid new model validations and tuning reports. His analogy: a water filter on the tap instead of ripping out all the pipes.
• L1 is going away, and that could be good news. Piatetsky expects L1 alert review to be fully automated and much of L2 to follow, freeing analysts to focus on alerts that matter and giving large institutions real visibility into what is happening inside them for the first time.

About Peter Piatetsky

Peter Piatetsky began his career as an intelligence analyst in Afghanistan, then joined the US Treasury Department, where he worked at OFAC on Iran and later became the lead Russia analyst. He went on to run US compliance for a major Asian bank before co-founding Castellum.AI in 2019. Castellum's clients include a top four US bank, a top three global payments provider and a top five global crypto exchange.

Connect with Fintech One-on-One:

• Tweet me @PeterRenton
• Connect with me on LinkedIn
• Find previous Fintech One-on-One episodes

Ilya Fridman is the founder and CEO of Byzfunder, a small business lender that started originating just as COVID hit in 2020 and is on track to fund more than $700 million this year with only 100 employees. In September Byzfunder launched TraceDataIQ, an AI underwriting platform that layers multiple large language models on top of bank, tax, background and public records data. We talk about how it was built, what it has done to performance and staffing, and why Ilya thinks AI will create a new wave of small business owners.

What We Covered

• From Morgan Stanley to mortgages to merchant cash advance
• Building one of the top independent sales offices in the country
• Why COVID was a blessing in disguise for a brand new lender
• Term loans, MCAs and a fast-growing revolving credit line
• The 60,000 small businesses Byzfunder sees every month
• Why off-the-shelf bank statement parsers fall short for small business lending
• Writing down every rule a human underwriter follows
• Running nine language models across the underwriting process
• Funding with just a Plaid link in under an hour
• How the human underwriter's job went from 10 minutes to two
• Byzfunder's first ABS deal and becoming more institutional
• Pricing transparency in revenue-based finance
• Why AI will produce more entrepreneurs

Key Takeaways

• Better data lets you lend deeper, not just safer. Byzfunder is not using AI to push its default rate down. It holds charge-off targets steady and uses more predictive models to approve more businesses at the same cost, with early 90-day-plus performance running about 250 basis points better than baseline.
• AI underwriting is mostly codified human judgment. The team wrote down every rule a human underwriter applies, from reading bank statements to spotting a mismatched name on a tax return, then trained a stack of roughly nine language models from different providers to apply them.
• Humans stay in the loop, but their job changes. Underwriters now get an instant summary of every risk flag instead of digging through 300-page background reports, cutting review time from about 10 minutes to two. Human oversight remains above a certain dollar amount because hallucinations still happen.
• Operating leverage is the real prize. Headcount is up around 15% year over year while originations are up 75% and units more than 150%, so Byzfunder can onboard more partners without sacrificing speed or service.

About Ilya Fridman

Ilya Fridman is the founder and CEO of Byzfunder. He began his career as an investment banker at Morgan Stanley before moving into residential mortgages, then built a merchant cash advance independent sales office that he sold to BFS Capital in 2014. He ran revenue at BFS Capital for about three and a half years before launching Byzfunder.

Connect with Fintech One-on-One:

• Tweet me @PeterRenton
• Connect with me on LinkedIn
• Find previous Fintech One-on-One episodes

Daniel Simon is the founder and CEO of Coast, which builds fleet payments, expense management and bill pay for businesses that work in the field, from plumbers and HVAC companies to construction firms and school bus operators. He previously co-founded Bread, an early buy now pay later company that sold for around $500 million in 2020. We talk about how Coast replaced the old fuel card model with open-loop Visa cards and real-time vehicle data, why going deep on the trades beats building for everyone, and how the AI data center build-out is driving demand across the physical economy.

What We Covered

• From Wall Street software engineer to Yale Law School
• The insight behind Bread and the early days of buy now pay later
• How essential workers during the pandemic led to fleet payments
• Why fintech left the physical economy behind
• The problems with closed-loop fuel cards
• Interchangeable cards and SMS check-ins instead of PIN pads
• Using telematics data to block fuel fraud in real time
• Expanding from fuel into expense management and bill pay
• Why Coast targets light and medium duty fleets, not over-the-road trucking
• The AI data center build-out and demand for skilled trades
• Going deep on vertical needs versus horizontal expense platforms
• Embedded partnerships and the five-year vision

Key Takeaways

• The phone in the driver's pocket changes everything. Legacy fuel cards rely on PINs and odometer readings typed into a pump, which drivers share and fudge. Coast sends interchangeable cards that drivers check into by text message, so every transaction is tied to the right person and the right vehicle.
• Vehicle data plus transaction data stops fraud. By merging telematics GPS and fuel gauge readings with line item data from Visa, Coast can decline a purchase when the vehicle is nowhere near the pump, or cap a fill-up at what the tank can actually hold.
• Fuel is the wedge, not the business. Customers spend 3 to 5% of revenue on fuel but 15 to 30% on things they already put on cards. That is why Coast has expanded into expense management and bill pay, with working capital, business banking and more on the roadmap.
• The AI boom has a physical side. Data centers need electricians, plumbers and HVAC crews, and every business in the trades is now competing for the same workers. That makes control and visibility over a growing field workforce more valuable than ever.

About Daniel Simon

Daniel Simon is the founder and CEO of Coast, which he started in 2020. He began his career as a software engineer at a boutique technology consultancy, leading teams that built systems for investment banks, asset managers and hedge funds. He then went to Yale Law School, where he co-founded Bread, a consumer credit and payments company that was sold in 2020 for around $500 million.

Connect with Fintech One-on-One:

• Tweet me @PeterRenton
• Connect with me on LinkedIn
• Find previous Fintech One-on-One episodes