--c77b343b06f9a010 x-next-cache-tags: _N_T_/layout,_N_T_/page,_N_T_/ vary: RSC, Next-Router-State-Tree, Next-Router-Prefetch Bhushan Songire - Portfolio
person peeking from behind laptop
Available for new projects

Hi, I'm Bhushan 👋

I specialize in building intelligent, agentic AI solutions. I build generative applications using LangChain and modern GenAI frameworks. My work spans from designing context-aware agents to developing data-driven ML models and visual analytics. I unite GenAI with practical insights.

About Me

A glimpse into my world

AI Icon
InsightAI
AI Icon

Ready to showcase my expertise...

Scroll to this section to begin

My Experience

Here are some of my professional projects, showcasing my skills in various domains and applications

2026

KPMG India

During my time at KPMG India, I contributed to a major enterprise digitalization initiative for a government sector client. My primary focus was on developing and customizing ERPNext modules using Python and Frappe to modernize recruitment and workforce management systems. To elevate the platform's capabilities, I integrated Generative AI automation, including an intelligent chatbot designed to streamline user interactions and query resolution directly within the site. By orchestrating automated multi-stage approval workflows and implementing strict role-based access controls, I helped drive public sector digitalization and significantly reduced manual operational turnaround times.

2025

TP InfoTech

During my internship at TP InfoTech, I had the opportunity to directly contribute to real client projects, applying my technical skills in a practical, results-driven environment. I worked extensively on Agentic AI solutions, integrating contextual data into predictive machine learning models and structured dataframes to enhance model performance. Additionally, I was involved in AWS-based deployment of applications using FastAPI, PHP, and MariaDB, gaining practical cloud infrastructure experience. I also developed interactive dashboards for data visualization, trend analysis, and strategic decision-making.

Hourly PredictionsDay-wise FootfallCalendar HeatmapMonth over Month revenue comparison

2024

Freelance Developer

As a freelance AI Engineer, I specialize in building and deploying custom, end-to-end full-stack solutions. A major focus of my work has been developing secure, offline-capable Retrieval-Augmented Generation (RAG) systems for client applications. Leveraging robust open-source foundations, I engineered a comprehensive document intelligence platform powered by DeepMind's Gemma-3, utilizing Ollama for local inference. I designed a modular backend with FastAPI, LangChain, and FAISS for efficient vector retrieval, paired with a responsive Streamlit interface featuring live LLM response streaming. By implementing secure authentication and containerizing the entire architecture with Docker, I delivered scalable, production-ready AI tools.

RAG user authenticationRAG file and embedding stepsRAG source citationsRAG reasoning support

LangChain

LangGraph

Ollama

OpenAI

HuggingFace

AWS

Numpy

Pandas

Matplotlib

OpenCV

Firebase

SQL

FastAPI

Streamlit

GitHub

Docker

Raspberry Pi

Linux

My Tech Universe

Tracing my learning curve through the evolving tools, frameworks, and stacks I’ve worked with.

Loading Tech Stack...

Real World Results

See how I transformed my raw concepts into real world projects

Top Projects

RAG with LangChain and Gemma-3

RAG with Gemma-3 is a fully local system that lets users upload documents (PDF, TXT, MD, etc.) and chat with them using natural language. It runs on Gemma-3 LLM and mxbai-embeddings via Ollama, ensuring privacy and low-latency responses without needing the cloud. Built using LangChain, FastAPI, Streamlit, FAISS, Docker, and SQLite, it supports multi-file uploads, vector embeddings, real-time streaming replies, user authentication, and session history. Designed to be modular, clean, and fast - perfect for personal assistants, or educational use.

  • LangChain
  • FastAPI
  • Streamlit
  • Docker
  • Gemma-3
  • Ollama
  • FAISS
  • GitHub Actions
  • Hugging Face Spaces
Deployment
GitHub Repo
Demo Video
RAG with LangChain and Gemma-3

Inquisitive: GenAI Powered Quiz Generator

This is an AI-powered multilingual quiz generator that builds interactive quizzes from just a topic or any detailed text like news article / guide / notes / document in any language. Utilizing 'Gemini Flash' model for fast question generation, platform supports MCQ, Multiple-Correct, True/False, and Numerical ans type questions. It also provides result evaluation and analysis.

  • Python
  • Streamlit
  • Google Gemini
Deployment
GitHub Repo
Demo Video
Inquisitive: GenAI Powered Quiz Generator

QnA ChatBot with LangChain

A LangChain-powered QnA ChatBot, built with Streamlit. Supports real-time provider and model switching across OpenAI, Groq, Google, and Ollama. It dynamically lists models based on the selected LLM provider and maintains session history using advanced LangChain components.

  • LangChain
  • Streamlit
  • OpenAI
  • Groq
  • Google Gemini
  • Ollama
Deployment
GitHub Repo
Demo Video
QnA ChatBot with LangChain

LocalGPT: A Local ChatGPT

LocalGPT is an offline clone of ChatGPT completely built from scratch. Uses open-source LLMs via Ollama and offers an interactive Streamlit-based interface. The tool supports conversation threads, real-time streaming responses, and multimodal inputs including vision models. It retains context per thread, allows model switching, and runs entirely without internet, making it ideal for local experimentation with LLMs.

  • Ollama
  • Streamlit
  • Python
  • LLMs
  • Vision Models
Deployment
GitHub Repo
Demo Video
LocalGPT: A Local ChatGPT

Smart Attendance System

The Smart Attendance System automates attendance calculation during lectures using video input. It processes frames using OpenCV and matches faces against pre-trained encodings. Attendance is based on 75% presence across frames. The UI is built with Flask server and provides downloadable Excel reports. Features include threading for concurrent processing, dark/light themes, and real-time feedback.

  • OpenCV
  • Flask
  • Face Recognition
  • Threading
  • HTML
  • CSS
  • JavaScript
Deployment
GitHub Repo
Demo Video
Smart Attendance System

Distributed Attendance System

The Distributed Attendance System extends my 'Smart Attendance System' by enabling distributed video processing across multiple client machines. It supports both static and dynamic load balancing for face recognition-based attendance marking. The web-based interface allows video uploads, and results are downloadable in Excel format. It ensures faster processing using distributed computing, all coordinated by a central Flask server.

  • Distributed Computing
  • Socket Programming
  • Flask
  • OpenCV
  • HTML
  • CSS
  • JS
Deployment
GitHub Repo
Demo Video
Distributed Attendance System

Air Quality Monitoring System with Emergency Alerts

This project is a real-time air quality and HVAC monitoring dashboard that collects data from various sensors for environmental factors via an Arduino and Raspberry Pi setup. It adheres to the Cloud-to-Fog-to-Things (C2F2T) model, ensuring low-latency, distributed processing. A local visualization is displayed on LCD-TFT screen. Also, The web app provides an interactive and responsive interface for real-time data visualization and emergency alerts, and uses Firebase for real-time data storage. Deployed on Vercel, it offers seamless access to both historical and live sensor data.

  • Raspberry Pi
  • Arduino
  • Firebase
  • Vercel
  • Python
  • HTML
  • CSS
  • JavaScript
Deployment
GitHub Repo
Demo Video
Air Quality Monitoring System with Emergency Alerts

Ms. Minutes: Voice Assistant and Home Automation

Ms Minutes is a standalone, autonomous voice assistant inspired by the Loki character. It responds to voice commands for weather, calculations, definitions, and general questions. The assistant also includes home automation capabilities, controlling smart appliances like a solenoid latch. Integrated with the OpenAI API and deployed on a Raspberry Pi Zero 2-W, the system functions independently and can be used anywhere with just a power adapter.

  • Raspberry Pi
  • Python
  • OpenAI API
  • Speech Recognition
  • TTS
  • Home Automation
Deployment
GitHub Repo
Demo Video
Ms. Minutes: Voice Assistant and Home Automation

Travel Guide on AWS - II

This project is incremental improvement of the previous travel guide application. Moving forward from the basic deployment, it now includes more AWS Services like RDS for database management, Cognito for user authentication, and SNS for notifications. The project is focused on utilizing multiple AWS services for scalability and security.

  • AWS
  • EC2
  • RDS
  • S3
  • EBS
  • Cognito
  • SNS
  • Translate
  • Flask
  • HTML
  • CSS
  • JavaScript
Deployment
GitHub Repo
Demo Video
Travel Guide on AWS - II

Generative AI

RAG with LangChain and Gemma-3

RAG with Gemma-3 is a fully local system that lets users upload documents (PDF, TXT, MD, etc.) and chat with them using natural language. It runs on Gemma-3 LLM and mxbai-embeddings via Ollama, ensuring privacy and low-latency responses without needing the cloud. Built using LangChain, FastAPI, Streamlit, FAISS, Docker, and SQLite, it supports multi-file uploads, vector embeddings, real-time streaming replies, user authentication, and session history. Designed to be modular, clean, and fast - perfect for personal assistants, or educational use.

  • LangChain
  • FastAPI
  • Streamlit
  • Docker
  • Gemma-3
  • Ollama
  • FAISS
  • GitHub Actions
  • Hugging Face Spaces
Deployment
GitHub Repo
Demo Video
RAG with LangChain and Gemma-3

Inquisitive: GenAI Powered Quiz Generator

This is an AI-powered multilingual quiz generator that builds interactive quizzes from just a topic or any detailed text like news article / guide / notes / document in any language. Utilizing 'Gemini Flash' model for fast question generation, platform supports MCQ, Multiple-Correct, True/False, and Numerical ans type questions. It also provides result evaluation and analysis.

  • Python
  • Streamlit
  • Google Gemini
Deployment
GitHub Repo
Demo Video
Inquisitive: GenAI Powered Quiz Generator

QnA ChatBot with LangChain

A LangChain-powered QnA ChatBot, built with Streamlit. Supports real-time provider and model switching across OpenAI, Groq, Google, and Ollama. It dynamically lists models based on the selected LLM provider and maintains session history using advanced LangChain components.

  • LangChain
  • Streamlit
  • OpenAI
  • Groq
  • Google Gemini
  • Ollama
Deployment
GitHub Repo
Demo Video
QnA ChatBot with LangChain

LocalGPT: A Local ChatGPT

LocalGPT is an offline clone of ChatGPT completely built from scratch. Uses open-source LLMs via Ollama and offers an interactive Streamlit-based interface. The tool supports conversation threads, real-time streaming responses, and multimodal inputs including vision models. It retains context per thread, allows model switching, and runs entirely without internet, making it ideal for local experimentation with LLMs.

  • Ollama
  • Streamlit
  • Python
  • LLMs
  • Vision Models
Deployment
GitHub Repo
Demo Video
LocalGPT: A Local ChatGPT

Computer Vision and Distributed Computing

Smart Attendance System

The Smart Attendance System automates attendance calculation during lectures using video input. It processes frames using OpenCV and matches faces against pre-trained encodings. Attendance is based on 75% presence across frames. The UI is built with Flask server and provides downloadable Excel reports. Features include threading for concurrent processing, dark/light themes, and real-time feedback.

  • OpenCV
  • Flask
  • Face Recognition
  • Threading
  • HTML
  • CSS
  • JavaScript
Deployment
GitHub Repo
Demo Video
Smart Attendance System

Distributed Attendance System

The Distributed Attendance System extends my 'Smart Attendance System' by enabling distributed video processing across multiple client machines. It supports both static and dynamic load balancing for face recognition-based attendance marking. The web-based interface allows video uploads, and results are downloadable in Excel format. It ensures faster processing using distributed computing, all coordinated by a central Flask server.

  • Distributed Computing
  • Socket Programming
  • Flask
  • OpenCV
  • HTML
  • CSS
  • JS
Deployment
GitHub Repo
Demo Video
Distributed Attendance System

IoT and Edge Computing

Air Quality Monitoring System with Emergency Alerts

This project is a real-time air quality and HVAC monitoring dashboard that collects data from various sensors for environmental factors via an Arduino and Raspberry Pi setup. It adheres to the Cloud-to-Fog-to-Things (C2F2T) model, ensuring low-latency, distributed processing. A local visualization is displayed on LCD-TFT screen. Also, The web app provides an interactive and responsive interface for real-time data visualization and emergency alerts, and uses Firebase for real-time data storage. Deployed on Vercel, it offers seamless access to both historical and live sensor data.

  • Raspberry Pi
  • Arduino
  • Firebase
  • Vercel
  • Python
  • HTML
  • CSS
  • JavaScript
Deployment
GitHub Repo
Demo Video
Air Quality Monitoring System with Emergency Alerts

Ms. Minutes: Voice Assistant and Home Automation

Ms Minutes is a standalone, autonomous voice assistant inspired by the Loki character. It responds to voice commands for weather, calculations, definitions, and general questions. The assistant also includes home automation capabilities, controlling smart appliances like a solenoid latch. Integrated with the OpenAI API and deployed on a Raspberry Pi Zero 2-W, the system functions independently and can be used anywhere with just a power adapter.

  • Raspberry Pi
  • Python
  • OpenAI API
  • Speech Recognition
  • TTS
  • Home Automation
Deployment
GitHub Repo
Demo Video
Ms. Minutes: Voice Assistant and Home Automation

Web-Dev and Cloud Projects

Air Quality Monitoring System with Emergency Alerts

This project is a real-time air quality and HVAC monitoring dashboard that collects data from various sensors for environmental factors via an Arduino and Raspberry Pi setup. It adheres to the Cloud-to-Fog-to-Things (C2F2T) model, ensuring low-latency, distributed processing. A local visualization is displayed on LCD-TFT screen. Also, The web app provides an interactive and responsive interface for real-time data visualization and emergency alerts, and uses Firebase for real-time data storage. Deployed on Vercel, it offers seamless access to both historical and live sensor data.

  • Raspberry Pi
  • Arduino
  • Firebase
  • Vercel
  • Python
  • HTML
  • CSS
  • JavaScript
Deployment
GitHub Repo
Demo Video
Air Quality Monitoring System with Emergency Alerts

Travel Guide on AWS

This project is a beginner-friendly travel guide application deployed fully on AWS infrastructure. It uses EC2 for hosting the Flask backend, SQLite for storing location data, S3 for public images, and presents travel recommendations through a clean web interface. The project demonstrates basic cloud deployment, database integration, and responsive web design, making it ideal for understanding AWS fundamentals.

  • AWS
  • EC2
  • S3
  • EBS
  • Flask
  • SQLite
  • HTML
  • CSS
  • JavaScript
Deployment
GitHub Repo
Demo Video
Travel Guide on AWS

Travel Guide on AWS - II

This project is incremental improvement of the previous travel guide application. Moving forward from the basic deployment, it now includes more AWS Services like RDS for database management, Cognito for user authentication, and SNS for notifications. The project is focused on utilizing multiple AWS services for scalability and security.

  • AWS
  • EC2
  • RDS
  • S3
  • EBS
  • Cognito
  • SNS
  • Translate
  • Flask
  • HTML
  • CSS
  • JavaScript
Deployment
GitHub Repo
Demo Video
Travel Guide on AWS - II

VTOP Extension

This simple Chrome extension enhances the VTOP (Chennai Campus) portal by dynamically adding a total marks row to each subject's marks table. Built with Manifest v3 and the Chrome Scripting API, it's lightweight, easy to install manually, and works seamlessly on Chromium-based browsers.

  • JavaScript
  • HTML
  • CSS
  • Chrome Extension
  • Manifest v3
  • Chrome Scripting API
Deployment
GitHub Repo
Demo Video
VTOP Extension

All Projects

RAG with LangChain and Gemma-3

RAG with Gemma-3 is a fully local system that lets users upload documents (PDF, TXT, MD, etc.) and chat with them using natural language. It runs on Gemma-3 LLM and mxbai-embeddings via Ollama, ensuring privacy and low-latency responses without needing the cloud. Built using LangChain, FastAPI, Streamlit, FAISS, Docker, and SQLite, it supports multi-file uploads, vector embeddings, real-time streaming replies, user authentication, and session history. Designed to be modular, clean, and fast - perfect for personal assistants, or educational use.

  • LangChain
  • FastAPI
  • Streamlit
  • Docker
  • Gemma-3
  • Ollama
  • FAISS
  • GitHub Actions
  • Hugging Face Spaces
Deployment
GitHub Repo
Demo Video
RAG with LangChain and Gemma-3

Inquisitive: GenAI Powered Quiz Generator

This is an AI-powered multilingual quiz generator that builds interactive quizzes from just a topic or any detailed text like news article / guide / notes / document in any language. Utilizing 'Gemini Flash' model for fast question generation, platform supports MCQ, Multiple-Correct, True/False, and Numerical ans type questions. It also provides result evaluation and analysis.

  • Python
  • Streamlit
  • Google Gemini
Deployment
GitHub Repo
Demo Video
Inquisitive: GenAI Powered Quiz Generator

QnA ChatBot with LangChain

A LangChain-powered QnA ChatBot, built with Streamlit. Supports real-time provider and model switching across OpenAI, Groq, Google, and Ollama. It dynamically lists models based on the selected LLM provider and maintains session history using advanced LangChain components.

  • LangChain
  • Streamlit
  • OpenAI
  • Groq
  • Google Gemini
  • Ollama
Deployment
GitHub Repo
Demo Video
QnA ChatBot with LangChain

LocalGPT: A Local ChatGPT

LocalGPT is an offline clone of ChatGPT completely built from scratch. Uses open-source LLMs via Ollama and offers an interactive Streamlit-based interface. The tool supports conversation threads, real-time streaming responses, and multimodal inputs including vision models. It retains context per thread, allows model switching, and runs entirely without internet, making it ideal for local experimentation with LLMs.

  • Ollama
  • Streamlit
  • Python
  • LLMs
  • Vision Models
Deployment
GitHub Repo
Demo Video
LocalGPT: A Local ChatGPT

Smart Attendance System

The Smart Attendance System automates attendance calculation during lectures using video input. It processes frames using OpenCV and matches faces against pre-trained encodings. Attendance is based on 75% presence across frames. The UI is built with Flask server and provides downloadable Excel reports. Features include threading for concurrent processing, dark/light themes, and real-time feedback.

  • OpenCV
  • Flask
  • Face Recognition
  • Threading
  • HTML
  • CSS
  • JavaScript
Deployment
GitHub Repo
Demo Video
Smart Attendance System

Distributed Attendance System

The Distributed Attendance System extends my 'Smart Attendance System' by enabling distributed video processing across multiple client machines. It supports both static and dynamic load balancing for face recognition-based attendance marking. The web-based interface allows video uploads, and results are downloadable in Excel format. It ensures faster processing using distributed computing, all coordinated by a central Flask server.

  • Distributed Computing
  • Socket Programming
  • Flask
  • OpenCV
  • HTML
  • CSS
  • JS
Deployment
GitHub Repo
Demo Video
Distributed Attendance System

Air Quality Monitoring System with Emergency Alerts

This project is a real-time air quality and HVAC monitoring dashboard that collects data from various sensors for environmental factors via an Arduino and Raspberry Pi setup. It adheres to the Cloud-to-Fog-to-Things (C2F2T) model, ensuring low-latency, distributed processing. A local visualization is displayed on LCD-TFT screen. Also, The web app provides an interactive and responsive interface for real-time data visualization and emergency alerts, and uses Firebase for real-time data storage. Deployed on Vercel, it offers seamless access to both historical and live sensor data.

  • Raspberry Pi
  • Arduino
  • Firebase
  • Vercel
  • Python
  • HTML
  • CSS
  • JavaScript
Deployment
GitHub Repo
Demo Video
Air Quality Monitoring System with Emergency Alerts

Ms. Minutes: Voice Assistant and Home Automation

Ms Minutes is a standalone, autonomous voice assistant inspired by the Loki character. It responds to voice commands for weather, calculations, definitions, and general questions. The assistant also includes home automation capabilities, controlling smart appliances like a solenoid latch. Integrated with the OpenAI API and deployed on a Raspberry Pi Zero 2-W, the system functions independently and can be used anywhere with just a power adapter.

  • Raspberry Pi
  • Python
  • OpenAI API
  • Speech Recognition
  • TTS
  • Home Automation
Deployment
GitHub Repo
Demo Video
Ms. Minutes: Voice Assistant and Home Automation

Travel Guide on AWS

This project is a beginner-friendly travel guide application deployed fully on AWS infrastructure. It uses EC2 for hosting the Flask backend, SQLite for storing location data, S3 for public images, and presents travel recommendations through a clean web interface. The project demonstrates basic cloud deployment, database integration, and responsive web design, making it ideal for understanding AWS fundamentals.

  • AWS
  • EC2
  • S3
  • EBS
  • Flask
  • SQLite
  • HTML
  • CSS
  • JavaScript
Deployment
GitHub Repo
Demo Video
Travel Guide on AWS

Travel Guide on AWS - II

This project is incremental improvement of the previous travel guide application. Moving forward from the basic deployment, it now includes more AWS Services like RDS for database management, Cognito for user authentication, and SNS for notifications. The project is focused on utilizing multiple AWS services for scalability and security.

  • AWS
  • EC2
  • RDS
  • S3
  • EBS
  • Cognito
  • SNS
  • Translate
  • Flask
  • HTML
  • CSS
  • JavaScript
Deployment
GitHub Repo
Demo Video
Travel Guide on AWS - II

VTOP Extension

This simple Chrome extension enhances the VTOP (Chennai Campus) portal by dynamically adding a total marks row to each subject's marks table. Built with Manifest v3 and the Chrome Scripting API, it's lightweight, easy to install manually, and works seamlessly on Chromium-based browsers.

  • JavaScript
  • HTML
  • CSS
  • Chrome Extension
  • Manifest v3
  • Chrome Scripting API
Deployment
GitHub Repo
Demo Video
VTOP Extension

Contact Me

My inbox is always open. Whether you have a question or just want to say hello, I'll try my best to get back to you! Feel free to mail me about any relevant job updates.

--c77b343b06f9a010--