Jatin Saini

An AI-powered resume analysis platform that evaluates resumes against job descriptions and generates structured feedback using LLM APIs.

• Architected an AI-driven analysis engine using Next.js and LLM APIs to compare resumes with job descriptions and produce structured JSON outputs.
• Developed a full-stack system with persistent analysis history and real-time UI updates using Firebase Firestore.
• Optimized prompt-driven workflows to identify keyword gaps and generate match scores for resume improvement.

A collaborative task management platform using Next.js, React, TypeScript, and Firebase with real-time updates and secure authentication.

• Implemented task assignment, workspace collaboration, and integrated a Node.js + Express backend with Firestore for scalable and efficient data handling.

A real-time speech translation application that converts Hindi audio to English text using advanced AI models.

• Built a React + TypeScript speech translation app using MediaRecorder to capture audio and Groq's Whisper API to translate Hindi speech to English text.
• Implemented async API handling with FormData uploads, error handling for API failures and microphone access, and state management for recording, loading, and translation states.

A browser extension that generates concise summaries of webpage content using the Google Gemini API.

• Built a browser extension using JavaScript that extracts webpage content and sends it to the Google Gemini API to generate concise summaries.
• Injected custom scripts into active tabs, handled DOM parsing, and implemented API integration for real-time, AI-powered summarization.

Engineered a meta-agent with the Google Agent Development Kit (ADK) that can dynamically generate executable Python scripts for specialized AI agents based on natural language descriptions.

• Demonstrated proficiency in agentic workflows by building a single agent that can perform a complex, multi-step task and also serve as a general conversational assistant.
• Showcased an understanding of multi-agent system principles and conditional logic, allowing the agent to dynamically switch between acting as a builder or a chatbot based on user intent.

A production-ready Retrieval-Augmented Generation (RAG) chatbot that allows users to upload PDFs and ask context-aware questions over private documents.

• Built a complete RAG pipeline using FastAPI, ChromaDB, and HuggingFace embeddings to ingest, chunk, embed, and retrieve document context.
• Implemented PDF upload, text extraction, normalization, and persistent vector storage with namespace-based retrieval.
• Integrated Groq LLMs for grounded answer generation, ensuring responses are strictly based on retrieved document chunks.
• Designed a Next.js dashboard with conditional chat logic to switch seamlessly between normal LLM chat and document-aware RAG responses.

A production-ready AI-powered REST API that analyzes resumes against job descriptions and returns structured insights.

• Built a scalable Node.js + TypeScript REST API using clean architecture, decoupling business logic from LLM providers via interfaces and factory patterns.
• Implemented secure API key authentication, per-key rate limiting, usage tracking, and expiry to simulate real-world SaaS API behavior.
• Integrated Groq LLM for resume analysis and deployed the service on Render with environment-based configuration.