Skip to main content
Back to challenges
Challenge

AI SDK & SambaNova Cloud UPI Fraud Triage Stream

High-volume digital brokers and fintech apps need ultra-low latency transaction fraud classification on high-throughput payment rails. You will construct a streaming agent that processes real-time device telemetry and transaction logs to flag anomalous money transfers with sub-50ms latency.

Business OperationsHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Build a Vercel AI SDK real-time agent powered by SambaNova Cloud fast inference to evaluate UPI transaction streams against device risk graphs.

Delivery guide

How work is evaluated

Evaluation

Evaluates streaming fraud triage latency and classification accuracy on UPI transaction payloads.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

Latency Threshold Test

Checks that the processing time remains below 50ms

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

Fraud Detection Precision

Precision of identifying fraudulent transfers • target: 0.97 • range: 0.92-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

What you should walk away with

  • Construct streaming agent pipelines using Vercel AI SDK streamText and tool calling

  • Integrate SambaNova Cloud ultra-fast inference API endpoints for sub-50ms response

  • Process real-time UPI payment signals, device fingerprints, and geolocation velocity

  • Implement continuous risk evaluation loops for dynamic payment blocking

Resources and assets

Reference links and supporting material

Dataset notes

Real-time simulated UPI transaction stream with embedded anomaly patterns such as velocity spikes and novel device logins.

Challenge setup

Evaluates streaming fraud triage latency and classification accuracy on UPI transaction payloads.

Challenge input

JSON transaction payload with sender, receiver, amount, and velocity score

AI SDK

Provides fast streaming tools for TypeScript/JS environments.

SambaNova Cloud

Delivers ultra-fast LLM inference for low latency requirements.

Evaluated output

JSON with action (ALLOW|BLOCK|FLAG), risk_score, and latency_ms

Evaluation checks
  • Checks that the processing time remains below 50ms
  • Precision of identifying fraudulent transfers • target: 0.97 • range: 0.92-1
Configured evaluation evidence
  • Benchmark: PaymentRisk-Bench
  • Fraud Detection Precision target: 0.97
  • 1 public reference case
Configured runtime
  • JavaScript execution harness
  • javascript sandbox (unavailable on Versalist)

Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.

View technical recipe

Configured tools

Action Space
  • SambaNova Cloud · Required
  • RAI · Optional
  • Synthflow · Optional

Evaluation contract

  • Latency Threshold Test · Weight 1
  • Fraud Detection Precision · Weight 1

Metrics

  • Fraud Detection Precision · Target 0.97

Benchmarks

  • PaymentRisk-Bench
  • security

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON transaction payload with sender, receiver, amount, and velocity score

    Output: JSON with action (ALLOW|BLOCK|FLAG), risk_score, and latency_ms

    Execution harness: JavaScript

Evaluation tools

  • AI SDK · Streaming Framework
  • SambaNova Cloud · Inference Acceleration

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about AI SDK & SambaNova Cloud UPI Fraud Triage Stream