I am a Senior React Native Engineer with 7+ years of software development experience, building and shipping mobile applications across banking, e-commerce, education, media, and consumer products.

My recent work focuses on combining strong mobile engineering foundations with GenAI and Agentic AI, including:

• Retrieval-Augmented Generation (RAG)
• AI agents, state, memory, and tool calling
• Model Context Protocol (MCP)
• LangChain and LangGraph workflows
• Structured outputs with Pydantic
• Hybrid retrieval, reranking, and vector search
• LLM evaluation for groundedness, relevance, completeness, and hallucination
• Gemini and local/open-source LLM integration

I enjoy turning complex requirements into products that are structured, scalable, explainable, and ready for real users.

An AI-powered civic operations platform for reporting and managing public issues such as potholes, garbage, water leakage, and broken streetlights.

What I built:

• WhatsApp-style citizen and admin AI assistants
• Custom MCP server exposing complaint-management tools
• AI admin briefings powered by Gemini
• SLA prioritization and overdue-issue detection
• Duplicate complaint detection
• Role-based complaint workflows
• Worker assignment, notifications, and audit logging
• Mobile-first issue reporting and administration experience

Tech: TypeScript · Node.js · React · MCP · Gemini · Supabase · Tool Calling · Agent Workflows

Repository and live demo links can be added here when public.

A conversational real-estate assistant that helps users discover properties and complete actions through WhatsApp.

Capabilities:

• Receives and replies to WhatsApp messages through Meta Webhooks
• Uses Gemini as the conversational reasoning layer
• Searches property inventory stored in SQLite
• Sends property photos directly on WhatsApp
• Maintains user sessions and selected-property context
• Captures leads and books site visits
• Provides a lightweight local admin dashboard

Tech: Python · FastAPI · Gemini · Meta WhatsApp Cloud API · SQLite · AI Agents · Tool Calling

Repository link can be added here when public.

A learning and experimentation project for building grounded AI assistants over course material.

What I explored:

• Document ingestion, chunking, embeddings, and vector search
• Dense and BM25 hybrid retrieval
• Reranking for improved context relevance
• Structured and schema-validated responses
• Local and open-source LLM experimentation
• Lightweight fine-tuning
• LLM evals for retrieval relevance, groundedness, completeness, and hallucination
• Base-model versus fine-tuned-model response comparison

Tech: Python · LangChain · LangGraph · Pydantic · Vector Databases · Embeddings · RAG · LLM Evals

Repository link can be added here when public.

I have delivered production mobile applications for organizations and products including:

• Five Talents Marketing Agency — Holy Reads
• LTIMindtree / Riyad Bank — Personal Loan, Credit Card, Auto Lease, and Mortgage journeys
• Christian Lingua — The Abba Kid App
• Apps Maven — E-commerce, healthcare, logistics, and lifestyle applications

My mobile engineering work includes:

• React Native architecture and modular application design
• TypeScript-first development
• Offline-first media and data experiences
• Video, audio, quizzes, subscriptions, payments, and analytics
• Authentication, MFA, and security-sensitive banking flows
• Performance optimization and crash reduction
• Native Android and iOS integrations
• Code reviews, mentoring, and engineering standards

Production Mobile Engineering
        +
GenAI and Agentic AI Systems
        +
Product Thinking and Clear Architecture

I can contribute to traditional mobile engineering roles, AI-enabled mobile products, and teams building practical agentic systems around real business workflows.

I am interested in building meaningful products at the intersection of mobile engineering, GenAI, and Agentic AI.

• LinkedIn: Prakunj Chaudhary
• Email: prakunjchaudhary@gmail.com
• Location: India