SardineAI Corp offers a fully integrated suite of fraud and compliance products, including customer due diligence, fraud prevention, risk scoring, investigations, transaction monitoring, and case management. The company aims to provide modern tools to fight sophisticated fraud attacks using machine learning and data intelligence.

Sardine

SardineAI Corp

Sardine is a leader in financial crime prevention. Using unparalleled device intelligence and behavior biometrics, Sardine applies machine learning to detect and stop fraud before it happens. The platform includes tools for identity verification, fraud prevention and investigation, AML monitoring, and case management. Over 250 companies use Sardine to prevent fake account creation, social engineering scams, account takeovers, bot attacks, payment fraud, and money laundering. For more information, visit www.sardine.ai.

Village Global

XYZ Venture Capital

overtime.vc

11.2 Capital

Anshu Sharma

Nyca Partners

Andreessen Horowitz

Experian Ventures

The General Partnership

Visa

Activant Capital

ConsenSys

Cross River Digital Ventures

ING Ventures

Eric Schmidt

In the next 10 years, most bank BSA officers will be data scientists or upskill to be data literate.

Typically a bank or Fintech BSA/AML officer has an operations background. They will implement processes and procedures that are well-understood and industry standard. This model will not last.

Why? 👇

The operations-led approach to BSA/AML isn’t effective. We know SAR volumes are increasing, but false positive rates can be anywhere from 90% to 99% in a worst case. 

Additionally, the operations-led approach uses ad-hoc and sample testing to discover false positives. This isn’t data-led and doesn’t show true effectiveness with statistical rigor.

👉 If every BSA/AML team complains about too much manual process and review, they need to learn data science as a skill.

👉 Instead we need to take an analytical and data-led approach to compliance problems

This means: 

✅ Real-time data dashboards and analytics of false positives. 
✅ Applied to 100% of the user population
✅ Identifying gaps in controls on the entire use base
✅ Consistently labelling data, testing new controls and tweaking thresholds against historical data to find the most effective program

#transactionmonitoring #aml #compliance

Real-time transaction monitoring is table stakes. It improves data privacy, prevents wires and payments from being “stuck in limbo,” and reduces manual work for compliance teams.

How? 👇

There are two worlds in financial services risk: 

Fraud and compliance (KYC/AML).

👉 Historically Fraud was more real-time, trying to detect possible fraud before a transaction to block it or add friction.

👉 AML was after a transaction, trying to create a complex pattern of transactions and activity that could suggest coordinated action by a crime ring or sanctioned entity. Understandably this sometimes takes days or weeks.

This created different processes, teams, and priorities. These teams operated in silos. 

🐟🐟🐟

The consequence of missing a fraudulent transaction is lost income, unhappy customers, and lower conversion at checkout (less revenue). Fixing this problem is a commercial decision.

The consequences of inadequate compliance can be fines and, in some cases, action from regulators or loss of ability to operate. Fixing this is existential.

🐟🐟🐟

It's the same data.

When you break these silos, both sides get better results.

Real-time FrAML ("Fraud and AML") detection has:

👉1) Privacy of data: Fraud cases get escalated to compliance teams to file for a SAR above a  threshold ($2000 in the US). If teams use different systems, this data passes through non-secure or non-privacy preserving channels, e.g., emails. 

👉 2) Managing high risk payments: Assume a customer wires-in a huge amount of money. Assume also you have a policy that above a $$ threshold of transactions requires additional due diligence and document collection for that customer risk. 

❌ Option A: Without real-time transaction monitoring, you are forced to receive the money. 

1. You have to email the customer to gather supporting documentation, and if the supporting documentation is insufficient, begin steps to offboard the customer. 
2. But what do you do with the new wire transfer? 
3. While figuring this out, the wire transfer is in limbo, and both the sender and receiver are frustrated that the money has gone "missing."

✅ Option B: With real-time Transaction Monitoring, your platform can decline the incoming wire transfer and return the funds back to the sender. 

➡ Separately, your compliance team can begin the process to ask for additional supporting documentation. 


🐟🐟🐟

There are cases where real-time AML wouldn't work, and in those cases, we *also enable batch decisions*. 

But this is still better when it's data-science-led.

E.g., looking at user aggregate transaction volumes in a calendar month and comparing it to aggregate volumes across other users with similar profiles. 

Complex SQL must still be created to aggregate transaction counts and volumes. 

TL;DR

If your systems are real-time capable and batch. You get the best of both. 🙌

Real-time transaction monitoring is a critical tool banks and fintechs must use to improve their processes.


False positive ratios in AML are too high. The future is shifting from an operational-led mindset to a data-led mindset to reduce manual work AND more efficiently manage suspicious activity.

False positives result from transaction monitoring and case management solutions that aren't data-led.

If for every 1 fraud or money laundering transaction caught, 5, 10, or 20 good customers are also caught, this is incredibly ineffective. 

🐟🐟🐟

The hardest thing about being in the risk squad is the constant balancing act. The trade-off between reducing false positives and managing down fraud and money laundering risk.

All risk problems are data science problems

Solving data science problems needs more data, more data discipline, and better testing. 

Day to day that looks like A/B testing rules, A/B testing machine learning models, and constantly evaluating performance Vs customer feedback and historic data.

🐟🐟🐟

We're passionate about is that world-class data science, and data engineering is a great problem-solving muscle for organizations to build. 

But it must have humans in the loop from the risk squad who can understand and contextualize the data.

We're you're data science nerds who speak risk and compliance. And we're an extension of your team. We've been there, we care about effectiveness and we'd love to show you what we're working on.

Deep fakes pass KYC liveness checks. We've found the most reliable way to identify deep fakes is based on the user's device and behavior.

In this session, we see "Matt Cruise" fully passing a liveness check, but in our dashboard detected a bot and emulator running and flagged it as very high risk.

Deep fakes are able to circumvent most "deep fake" detectors, but the thing they can't get around is the lack of human behavior on the device.

Soups Ranjan will be sharing more about this at upcoming conferences like MRC | Merchant Risk Council - If you're keen to learn more, drop us a reply below :)

#kyc #deepfakes #deepfaketechnology 

Do you know why a payment failed? 

Was it an error or fraud? Collaborating with PSPs can help figure out the root cause and increase conversion, but still most merchants don’t do it.

🐟🐟🐟

You’ve been there. 

You’re trying to buy something online and it fails. 

You try again and it fails. 

Despite being a good customer you’re probably triggering a fraud rule at a merchant’s payments provider. 

🐟🐟🐟


The default setup for fraud rules is always a blunt instrument. It’s one-size-fits-nobody.

Good PSPs know this and are usually willing to work with their merchants to help tweak or change their default in certain situations. 

That could mean pushing fraud decisions to 3rd parties (like Sardine), or changing rules for that merchant.

🐟🐟🐟

Failed payments are expensive for merchants and frustrating for customers.

There’s so much value left on the table not chasing this down.

We had one client say “chasing payment performance makes basis point improvements in conversion, chasing fraud performance makes percentage points.” 

If you’re in payments, your best bang for buck is in fraud. 


AI:SARDINE

SardineAI Corp

locale

market

industry

website

crunchbase

linkedin

founding date

competitor of

Metrics