Nisarg  ·  Senior Python / AI Full Stack Developer  ·  5+ yrs

Overview: HRChatBot is an AI-powered interview assistant that automatically asks role-specific questions to interviewees. Responsibilities: Designed and developed an AI chatbot capable of conducting automated interviews based on user-selected roles. Integrated the Facebook classifier model for accurate intent detection and dynamic question flow. Implemented FastAPI backend for handling conversation logic and responses efficiently. Built the frontend using Python templates with HTML, CSS, and JavaScript for smooth interaction.

Overview: AskGuruji is an AI-powered spiritual knowledge chatbot designed to answer user queries using teachings from Guruji’s pravachan videos and related books. Responsibilities: Developed a system to extract audio and text from thousands of Guruji’s pravachan videos using FFmpeg and yt-dlp. Implemented chunking logic to divide transcripts and book content into semantically searchable segments. Integrated Pinecone for efficient vector storage and similarity-based retrieval. Utilized the Gemini API to generate precise, context-aware responses from relevant chunks. Built a React.js frontend for users to interact seamlessly with the chatbot.

Overview: Self Coder is an advanced AI-powered code generation platform where users simply describe their project aim, and the system automatically writes complete code, organizes it into corresponding files, and returns a downloadable ZIP file. Responsibilities: Designed and implemented an AI-based system capable of generating full project code from user descriptions. Integrated CodeLlama and LLaMA 2 models for intelligent code generation and text understanding. Built a FastAPI backend to handle requests, model processing, file generation, and ZIP packaging. Developed a React.js frontend for users to input project goals and download generated code easily.

Overview: Solar Detection is an AI-driven project aimed at identifying solar panels from satellite imagery. Responsibilities: Collected and curated a dataset of solar panel satellite images for model training. Fine-tuned the YOLOv8 model to accurately detect solar panels in various geographic and lighting conditions. Implemented a location extraction pipeline to pinpoint coordinates of detected panels. Integrated geolocation APIs to retrieve property and owner information from identified locations. Conducted testing and performance evaluation to ensure detection precision and low false positives.