Kavish Jaiswal

Solo-built AI-powered market intelligence platform targeting commercial deployment across Indian equities, global crypto, US markets, and commodities — all priced in USD and INR. Architected a multi-layer automated pipeline: Layer 1 aggregates live market data (CoinGecko, Alpha Vantage, NSE Direct, GoldAPI) via n8n using a cascading merge strategy to work around n8n's 10-input node limit; Layer 2 uses an LLM-based parser via OpenRouter with strict anti-hallucination guardrails, delivering daily reports as formatted HTML email and Discord embeds with a custom markdown-to-HTML renderer. Layer 3 (live): an LLM Council where GPT-4o-mini, Gemini Flash, and Claude Sonnet serve as independent analysts with Claude Opus as chairman synthesising a final report via anonymised peer review. FastAPI backend containerised on Railway, called from n8n via authenticated webhooks. In progress: Layer 4 RAG pipeline (Pinecone), Layer 5 conversational Next.js frontend with OpenAI Whisper STT and TTS, and multi-user scaling via Supabase replacing Google Sheets.

Architecting and building a custom end-to-end vacation property booking and management platform for two residential bungalows in Thane, Maharashtra. Full backend (Node/Express/TypeScript/MongoDB) with availability conflict prevention, manual UPI payment flow, security deposit tracking, AES-encrypted Aadhaar ID collection, audit logging, and role-based access control. Google Calendar API sync for owner remote access; Resend for transactional email notifications. Customer-facing frontend with live availability calendar, dynamic pricing calculator, and multi-bungalow selection for events. Separate mobile-first admin dashboard built for an elderly low-vision user — large fonts, large tap targets, custom-built AdminCalendar component after react-day-picker compatibility failures. Implemented DPDP Act 2023 compliance from day one: explicit consent, data minimization, 30-day TTL-based automated deletion, and full audit trails. Razorpay integration scoped and preserved but intentionally disabled pending GST registration — designed for seamless reactivation. Orchestrator-first development workflow with prompt files authored for Claude Code CLI and Cursor Agent to maintain codebase consistency.

Technical Lead & Full-Stack Engineer on an AI-powered goal and task management app built during the Outskill AI Fellowship (C3 Team 8). Designed the full-stack monorepo architecture (React 19 + TanStack Start + FastAPI + LangGraph), implemented the GoalPlannerAgent, Scheduler Agent, and Classifier Agent from scratch, and built an end-to-end RAG pipeline using Pinecone and text-embedding-3-small with source citations surfaced in the UI. Drove all 24 integration tests to green by diagnosing JWT HS256/ES256 mismatches, auth sync gaps, and route contract divergences. Resolved a critical silent failure where VITE_USE_MOCK=true masked all backend calls. Managed Git branch strategy across a 5-person team using archival tags to preserve contribution history.

Comprehensive AI-powered mobile fitness application combining workout tracking, nutrition analysis, and personalized coaching. Integrated computer vision for meal analysis, LLM-powered workout recommendations, and real-time activity monitoring. Built with React Native cross-platform architecture, Supabase backend, and OpenRouter AI APIs. Features advanced progress analytics, conversational AI assistant, and secure authentication with row-level security. Delivered seamless user experience across iOS, Android, and web platforms.

Complete overhaul of a fintech platform serving customers seeking loans and credit cards. Transformed from legacy architecture to modern, scalable solution with 35% performance improvement and enhanced user experience. Built with React.js frontend and Node.js backend.

Advanced trading bot that tracks, notifies and analyzes prices of major cryptocurrencies (BTC, ETH, BNB, SOL) in both INR and USD. Features real-time price monitoring, automated notifications, and comprehensive market analysis with Python backend.

Comprehensive Python API example demonstrating best practices in API development, including proper error handling, authentication, and documentation. Showcases clean code architecture and modern Python development techniques.

Intensive hands-on fellowship building production-ready AI applications with focus on agentic workflows and retrieval-augmented generation (RAG) systems. Currently implementing RAG pipeline for intelligent query answering using LangChain and Pinecone.