Where developers grow together

ResolveAI — Autonomous Fintech Customer-Service Platform

From Complaint to Verified Resolution ResolveAI is a fintech customer-service prototype that demonstrates how payment-related complaints can move through a structured, transparent resolution workflow. The Problem Failed payments, gateway timeouts, delayed transactions, and suspicious payment activity can create frustrating customer experiences. Resolving these cases requires organized investigation, clear decisions, and appropriate follow-up. Our Solution ResolveAI demonstrates a structured workflow that takes a complaint through five key stages: intake, investigation, decision-making, simulated action execution, and final verification or escalation. Key Features Structured Case Management: Organizes payment complaints into trackable cases. Multi-Stage Workflow: Visualizes the progression from complaint intake to resolution. Simulated Payment Scenarios: Demonstrates different transaction issues through predefined cases. Decision and Action Tracking: Makes the intended resolution process easier to understand. Verification and Escalation: Represents successful verification and alternative outcomes when a case cannot be resolved. Technology Built as a full-stack web application, ResolveAI combines a modern web interface with a structured workflow engine to demonstrate fintech customer-service operations. What Makes It Different? Rather than presenting customer support as a simple chatbot conversation, ResolveAI focuses on the complete complaint lifecycle, including investigation, action tracking, and verification. Project Links Live Demo: https://resolveai-i6ba.onrender.com/ Source Code: https://github.com/YuvrajGora/ResolveAI Important Note ResolveAI is an independent demonstration project that uses simulated payment scenarios. It does not process real financial transactions and is not officially affiliated with Paytm or any payment provider.

HR Management System

Human Resource Management System (HRMS) A Human Resource Management System (HRMS) is a web-based application designed to simplify and automate an organization's HR operations. It provides a centralized platform for managing employee information, attendance, leave requests, payroll, recruitment, and performance evaluations. The system enables HR administrators to maintain employee records, track attendance, manage leave approvals, generate payroll reports, and monitor employee performance efficiently. Employees can access their profiles, apply for leave, view attendance records, and check payroll information through a user-friendly interface. The HRMS improves accuracy, reduces manual paperwork, streamlines administrative processes, and enhances communication between employees and HR personnel. It helps organizations manage their workforce effectively while ensuring secure access to employee data through role-based authentication and authorization.

OMNI-SWARM-Speak in your language. Get verified help you can own.

OMNI-SWARM: Speak in your language. Get verified help you can own.   THE PROBLEM Many students, villagers and informal workers struggle with digital services. Forms are long, apps are many, and most tools are built for English speakers. People also have no trusted proof of their identity, skills or entitlements that they themselves own. On top of that, AI assistants are rarely checked for attacks like prompt injection before they act.   OUR SOLUTION OMNI-SWARM is a governed voice and text AI assistant for citizens. A person speaks or types a request in Kannada, Hindi, Tamil, Telugu or English. The platform protects their personal data, screens the request for attacks, helps with documents, government schemes, facilities, letters and reports, and can issue a signed record that the citizen owns and anyone can verify.   HOW IT WORKS 1. The user signs in and asks for help by voice or text, for example "How do I apply for a ration card?" 2. Aadhaar, phone, PAN and email are redacted before anything reaches the AI. 3. The security layer screens the input. Prompt-injection attempts are blocked and logged. 4. The assistant replies in the user's language with simple steps. It does not invent scheme names, amounts or deadlines, and it says where to verify. 5. For sensitive actions, a liveness challenge asks the user to speak a random phrase. 6. Once every check has passed, a credential is signed and the user downloads a signed Word document with a drawn signature and a QR code. 7. Anyone can scan the QR code and see Valid or Tampered on a public verification page.   KEY FEATURES - Assistant: multilingual voice and text chat, privacy redaction, document and scheme help - Bharat Agent: our Streamlit app embedded in the platform, clearly labeled as external - Agent Builder: Architect, Developer and QA agents that generate a small app with real tests, a sandboxed preview and a ZIP export - Security Center: prompt-injection guard, challenge-response liveness check, rate limiting, signed security certificates - Credential Wallet: W3C-style credentials signed with Ed25519, QR verification, hand-drawn signature, signed .docx export - Vendor Guardian: software bill of materials scanned against the OSV.dev vulnerability database, plus a third-party service list - Dashboard and Audit Log: live numbers from the database and a full, filterable log of every action   TRACK FIT (WEB3 & CYBERSECURITY, PS32) Identity and trust come from citizen-owned credentials that anyone can verify with public-key signatures. Ownership means the person holds the proof. On the cybersecurity side we cover prompt-injection defense, liveness checks, signed audit certificates and supply-chain risk scanning.   TECHNOLOGY React, TypeScript, Tailwind CSS, Supabase with edge functions and row-level security, Web Speech API with optional Sarvam or BHASHINI voice, an LLM through an OpenAI-compatible endpoint, Web Crypto SHA-256 and Ed25519 signatures, OSV.dev, and the docx and qrcode libraries.   WHAT MAKES IT DIFFERENT - Voice AI, security and verifiable credentials work together in one flow - Security runs before the AI acts, not after - Every number on screen comes from real data - Signed records can be checked by anyone, and tampering is detectable   HONEST LIMITS Browser speech recognition may use the browser vendor's servers. The liveness check is challenge-response, not a deepfake detector. The credential proof is a simplified Ed25519 signature. A signature drawn in a Word file is a visual mark; the real proof is the hash and signature checked on the verify page. The Agent Builder exports code and does not host live backends.   IMPACT Lower barriers for non-English speakers and first-time digital users, portable proof that the citizen owns, and safer use of AI in public-service tasks. We will measure the injection block rate on our own test set and the time from request to signed record.   WHAT'S NEXT More languages, official Sarvam and BHASHINI voice, anchoring credential hashes on a blockchain testnet, and pilots with schools, NGOs or welfare programmes.   TEAM Prism Talons, led by Prajval.