Hackathon Portal
AI Tinkerers - Hyderabad
Team

Celestial

Project Concept

No description has been added yet.

Entry

Status: Submitted

Last saved: May 09 at 5:36 PM IST

Team Roster

Message board not available for this team yet.

Kasula Manoj Team Lead RSVP Approved

Ai engineer at Antz ai
End-to-end build (sole contributor). - Architecture: Next.js 16 App Router + CopilotKit runtime adapter wired to Vertex AI Gemini 2.5 Flash via LangChain (`@langchain/google-vertexai`). Designed the "no pages" pattern — agent owns the canvas, widgets are rendered as side-effects of `useCopilotAction` handlers. - Generative UI layer (CopilotKit): authored 7 `useCopilotAction` tools — renderEmiCalculator, renderCibilSimulator, renderHardshipQuiz, renderDocUploader, renderPayoffViewer, renderSuccessorWizard, closeWorkspace — each materializing a distinct interactive React component on the LiveWorkspace canvas based on agent-emitted arguments. - Widgets built: EmiCalculator (live sliders, prepay savings), CibilSimulator (score drop + recovery curve, Recharts), HardshipQuiz (RBI moratorium eligibility), DocUploader (drag-drop → Google Document AI → field extraction), PayoffViewer (foreclosure cost + 80C/24b tax benefit), SuccessorWizard (death/divorce/transfer flows with document checklist). - Voice loop: bilingual EN/HI input via Google Cloud Speech-to-Text → Gemini → response routed through Cloud Text-to-Speech → 3D speaking avatar (Three.js + React Three Fiber) with state-driven mouth/eye animation and live captions. - API routes: /api/copilotkit (CopilotKit runtime), /api/parse (Document AI), /api/speech (STT), /api/tts (TTS), /api/translate (HI↔EN), /api/health. - Deploy: Dockerfile + deploy.ps1 → Google Cloud Run (asia-south1) with metadata-server auth. - UX: bilingual suggestion chips, Indian-rupee formatting, sticky borrower summary cards, voice/agent state indicators.
Kasula Manoj is an AI Engineer at Antz ai. His LinkedIn profile is https://www.linkedin.com/in/manoj-kasula-a131301b7.
Developing AI systems like Call-System (Twilio/Whisper/Gemini), Alternative-TTS, and Animated-Video-Generator. Utilizing FastAPI, gRPC, and LangChain for production-adjacent workflows. Managing the popular doom-coding repository while exploring AI agents.