# Pratik Jadhav | Full Stack Engineer

> Full Stack Engineer with 1+ year of production experience building scalable web applications and AI-powered products. Strong in React, Next.js, Node.js, MongoDB, Redis, Docker, RAG, and Agentic AI, with end-to-end ownership of production systems.

## Contact
- Email: pratikjadhav1438@gmail.com
- Phone: +91-8010316078
- Location: Pune, Maharashtra
- Resume: https://drive.google.com/drive/folders/1J4FDjdY7o_RPVfqac3Qle8XkEGp5OOTX
- GitHub: https://github.com/Pratik4230
- LinkedIn: https://linkedin.com/in/pratikjadhav1438
- X: https://x.com/Pratik4230
- Website: https://portray.vercel.app

## Skills
Frontend: React.js, Next.js, TypeScript, JavaScript, Tailwind CSS, TanStack Query, Zustand, Redux Toolkit
Backend: Node.js, Express.js, REST APIs, MongoDB, PostgreSQL, Redis, BullMQ, Socket.IO
AI & LLM: OpenAI SDK, Vercel AI SDK, LangGraph, LangChain, RAG, Agentic AI, Tool Calling, Sarvam AI
DevOps: Docker, Docker Swarm, GitHub Actions, CI/CD, Nginx, Cloudflare Workers

## Experience
### Software Developer at Technfest IT Solution (Jan 2025 to May 2026)
Stack: React.js · Node.js · MongoDB · Redis · Docker · BullMQ · Socket.IO
  - Owned development of a production multi-tenant ERP with RBAC across 10+ business modules, 30+ analytics reports, and 90+ MongoDB collections, supporting 3,000+ customer visits/day during peak periods
  - Cut master-data search latency by 67% (55ms → 18ms) by moving frequent searches from MongoDB to Redis Stack with RediSearch indexes, reducing database queries
  - Streamlined stock transfer processing by ~60%, enabling faster warehouse operations through real-time inventory synchronization
  - Reduced Purchase Invoice interaction latency by ~45% under heavy item loads by optimizing React rendering and decomposing a monolithic form
  - Engineered production infrastructure with Docker Swarm, GitHub Actions, MongoDB Replica Set, BullMQ, and Nginx, enabling automated zero-downtime releases

## Projects (Top 3)
### AiXpense
AI-powered expense tracker for India and beyond
AiXpense replaces expense forms with a chat-first experience. Say "Zomato 450" or "petrol 800 aaj" and the AI logs amount, category, and date automatically. Supports Hindi, Marathi, Hinglish, and 22+ Indian languages on web and Android, with budgets, recurring rules, natural-language analytics, and premium receipt scanning.
Links: Live: https://aixpense.in | Android: https://play.google.com/store/apps/details?id=com.aixpense.app | GitHub: https://github.com/Pratik4230/AiXpense
Tech: Next.js, TypeScript, Node.js, MongoDB, React Native, Expo, Vercel AI SDK, OpenAI, Sarvam AI, Inngest, Better Auth, TanStack Query
Features:
  - Natural language logging: One sentence logs expenses and income. The AI infers merchant, amount, category, tags, and date from context like "Swiggy lunch 320" or "received salary 50000".
  - Multilingual voice input: Speak in Hindi, Marathi, Hinglish, English, Tamil, Telugu, and 22+ Indian languages via Sarvam AI. Up to 26 seconds per voice note, transcribed server-side.
  - Bill and receipt scanning: Premium feature: snap a receipt in chat and the AI extracts merchant, total, date, and line items. Confirm and save in seconds without manual typing.
  - Smart categorization: Auto-assigns categories from a rich taxonomy (Food, Transport, Rent, EMI, Salary, Freelance, etc.) inferred from merchant names and wording.
  - Budget tracking: Set monthly budgets per category via chat or the Budgets screen. Warnings surface before overspend, not after the month ends.
  - Recurring payments: Premium rules automate rent, EMIs, subscriptions, and gym memberships. Scheduled spends drop into the ledger on the right dates.
  - Reports and analytics: Overview cards, category breakdowns, trend charts, budget vs actual, and top expenses on web and mobile Reports screens.
  - AI spending coach: Premium Inngest jobs send weekly and monthly coach emails with personalized insights and one actionable tip from real transaction data.
  - Offline queue (Android): Expense text, voice, and receipt photos queue locally when offline and sync FIFO when network returns. Share intent routes bill photos from any app.
  - Biometric app lock: Fingerprint or face unlock on Android adds a local security layer on top of Better Auth login, re-locking after background.

### Xaply
Short links that move fast
Xaply turns long URLs into go.xaply.in links that resolve in under 10ms from 300+ edge locations. The web dashboard and native Android app offer custom slugs, password-protected links, QR codes, smart platform redirects, team workspaces, webhooks, and click analytics with redirects kept off the control-plane path.
Links: Live: https://xaply.in | Android: https://play.google.com/store/apps/details?id=com.pratik4230.xaply | GitHub: https://github.com/Pratik4230/Zap
Tech: Next.js, Cloudflare Workers, Cloudflare D1, Cloudflare KV, Cloudflare Queues, Expo, React Native, Jetpack Compose, Better Auth, Drizzle ORM, TanStack Query
Features:
  - Edge-fast redirects: Cloudflare Workers resolve hot slugs from KV cache with automatic HTTPS. No spinner pages, no cold starts.
  - Link management: Custom slugs, edit anytime, pause/resume, expiry by date or click count, safe delete, server-side search, and infinite scroll.
  - Password protection: Hashed passwords gate sensitive links before redirect with rate-limited guess protection.
  - QR codes: Generate QR codes in the browser for posters, packaging, or events. Nothing uploaded or stored on servers.
  - Smart platform redirects: One short link routes Android, iOS, and default visitors to the right destination for app launch campaigns.
  - Real-time analytics: Global and per-link dashboards with countries, cities, devices, browsers, OS, and referrer breakdowns.
  - Workspaces and teams: Business plan supports up to 30 owned workspaces, 50 team seats, invites, and role-based access.
  - Webhooks: HMAC-SHA256 signed payloads for link.created, link.updated, link.deleted, and link.clicked with retry logic.
  - Xaply Mobile (Android): Native Material 3 app with link CRUD, analytics charts, workspace management, biometric lock, and push notifications for click milestones.
  - Streak rewards: 21-day app open streak grants 1 year of Pro free on the Android app.

### Manthan
Your documents. Your sources. Answers you can trust.
Manthan is a source-grounded AI research workspace. Users create notebooks, upload PDFs, DOCX, spreadsheets, web pages, and YouTube videos, then chat with an AI that answers only from those materials with inline citation chips, multi-thread history, and auto-generated briefings, FAQs, flashcards, quizzes, and mind maps.
Links: Live: https://manthan-ivory.vercel.app | GitHub: https://github.com/Pratik4230/Manthan
Tech: Next.js, TypeScript, LangGraph, LangChain, OpenAI, MongoDB Atlas Vector Search, Inngest, Better Auth, ImageKit, Firecrawl, assistant-ui, Turborepo
Features:
  - Source-grounded chat with citations: Answers come only from enabled, ready sources. Citation chips show source title, location (Page 4, Sheet: Revenue), and excerpt. If the answer is not in your materials, Manthan says so.
  - Multi-format source ingestion: PDF, DOCX, XLSX, CSV, TXT, MD via ImageKit. Web URLs scraped via Firecrawl. YouTube transcripts across all URL formats.
  - Real-time ingest progress: Granular stages: Queued, Downloading, Parsing, Chunking, Embedding, Summarizing, Indexing. Failed ingests retry; stale vectors re-index.
  - Smart duplicate detection: SHA-256 file hashing, normalized URL comparison, and YouTube videoId extraction with open, replace, or add-anyway options.
  - Intelligent query routing: LLM route step skips retrieval for greetings while document questions run full RAG with query rewrite, HyDE, and decomposition.
  - NotebookLM-style three-pane UI: Sources, Chat, and Studio panes with independent collapse, persisted state, and responsive flex sizing.
  - Multi-thread chat: Create, rename, archive, and delete threads with auto-derived titles and full message history in MongoDB.
  - Per-source control: Enable or disable individual sources mid-conversation to narrow or broaden what the AI can see.
  - Workspace instructions: Optional persona per workspace (e.g. legal analyst) injected into the system prompt without breaking source grounding.
  - Studio artifacts: Generate briefings, FAQs, flashcards, quizzes, and mind maps from workspace vectors via Inngest background jobs and OpenUI rendering.

## Education
B.Sc. Computer Science, Shivaji University
Yashwantrao Chavan College of Science, Karad, Maharashtra
Jun 2021 to Apr 2024

---
Canonical URL: https://portray.vercel.app
Last updated: 2026-09-01