razorpay-buildathon

AI risk manager for chargeback disputes — Razorpay AI Buildathon 2026, Track 2

Detects distributed card-testing campaigns in the region where per-entity velocity rules cannot fire at any threshold. 10 false positives against a baseline's 476.

Agent-2-Agent E-Commrce Platform [MVP Concept]

Multi-source financial reconciliation that matches what’s certain and honestly reasons about what isn’t. Track 04 for the Razorpay Buildathon

Autonomous AI revenue recovery and statutory compliance agent for Razorpay. Intelligently triages payment failures, eliminates card-network penalties, and coordinates multi-channel recovery with an immutable audit ledger.

Autonomous 4-Way Multi-Tier Settlement & Financial Reconciliation System with Zero-Drift Invariants, Monte Carlo Cash Forecaster & CFO Copilot. Built for Razorpay AI Buildathon 2026 (Track 04: AI Finance Controller).

𝑳𝒆𝒂𝒌𝑮𝒖𝒂𝒓𝒅 - an AI-powered system that turns failed/degraded payments into a controlled revenue-recovery process. 𝑳𝒆𝒂𝒌𝑮𝒖𝒂𝒓𝒅 - diagnoses the failure, evaluates actionability and recovery economics, selects the best intervention, enforces merchant policies and safety limits, executes it through providers like Razorpay/Twilio/Resend

Razorpay Buildathon Submission for Track 4: AI Finance Controller

Regime-aware fraud detection for merchants — built for the Razorpay Buildathon, Track 2: AI Risk Manager

AI-powered fraud ring detection agent combining ML risk scoring with graph-based relationship analysis. Built for the Razorpay AI Buildathon 2026 — AI Risk Manager track.

Agents can finally pay. The checkout AI buyers transact on, and the desk merchants trust — signed mandates, a hash-chained ledger, real Razorpay test rails. Track 1 · live at customs.srivtx.xyz

Autonomous multi-agent payment reliability & incident investigator platform. Resolves false failures, double debits, and bank switch outages using 11 deterministic engines + 4 collaborative AI agents. Built for Razorpay AI Buildathon 2026.

Autonomous, policy-bounded revenue recovery platform for Indian fintech and SaaS. Recovers failed payments via Expected Recovery Value (ERV) optimization, deterministic guardrails, and SHA-256 audit chaining. Razorpay AI Buildathon 2026 (Track 03).

AI finance controller that shows its work - bidirectional settlement reconciler with computed confidence, math-gated groups, and honest exceptions (Razorpay Buildathon Track 4)

AI finance controller for settlement-to-cash close with deterministic reconciliation, bounded AI investigation, human authority, and measured batch coverage.

Matches incoming bank credits to open invoices automatically. Handles TDS, lump sums, partial payments and mangled payer names. Cuts analyst time 4.00 to 1.59 min/credit at 99.7% precision. Razorpay AI Buildathon Track 04.

ASSAY — a bounded AI finance controller that reconciles evidence, abstains under ambiguity, and verifies ledger outcomes.

प्रमाण — Proof-of-Delegation & Dispute-Evidence Layer for AI-agent payments on Razorpay (test mode). The trust layer agentic commerce is missing: captured at delegation, weaponized at dispute, freeing authorized agents from false declines.

Merchant-side card-testing detection that shows its reasoning. Rules produce evidence, five explanations compete over it, and the model can ask for a review but never blocks a shopper — Razorpay Buildathon 2026, Track 02.

Revenue-integrity layer that plugs in under Razorpay — recovers failed payments with calibrated ML inside a bounded policy engine, and proves the incremental ₹ against a live randomized control. Measured, not estimated.