All Projects
Looksy
Status: Production
Thesis: Production AI portrait generation app — live on Google Play Store.
Looksy is a production AI portrait generation Full Stack Development mobile application built with React Native, TypeScript, and Firebase. It integrates Gemini 2.5 Flash for multimodal Image generation. Featuring Razorpay subscription billing, it leverages Next.js patterns for its landing pages and advanced Cloud Functions to guarantee secure Coin Ledger transactions.
- React Native
- Expo
- Firebase
- Gemini 2.5 Flash
- Razorpay
- TypeScript
FutureYogi AI
Status: Production
Thesis: Hast Rekha — AI palmistry platform with real-time PDF delivery.
FutureYogi AI is a monetised AI palmistry Full Stack Development platform leveraging Machine Learning via GPT-4o-mini vision inference to generate detailed palmistry dossiers. The architecture seamlessly converts reports to PDF on Cloud Run, optimized for the Indian market with robust Razorpay payment fraud prevention. Integrates beautifully with a modern Next.js and TypeScript web ecosystem.
- React Native
- Expo
- Python
- Flask
- GPT-4o-mini
- Cloud Run
- Firestore
- Razorpay
Chakshu
Status: Production
Thesis: AI-powered campus security — face recognition, weapons, fight detection.
Chakshu is an advanced Computer Vision campus security platform built with Deep Learning models including YOLOv8 and YOLOv11 for facial recognition, weapon detection, and fight detection on live CCTV streams. Deployed on NVIDIA Jetson Orin NX and optimized with DeepStream, the system utilizes a modern Next.js and React frontend with a Node.js backend to provide real-time alerts.
- YOLOv8
- YOLOv11
- OpenCV
- FastAPI
- React
- Node.js
- MongoDB
Automated Waste Segregation
Status: Patent
Thesis: Patented deep learning + IoT sorter for recycling optimisation.
An automated waste segregation system powered by Deep Learning and IoT. This patented technology leverages advanced Computer Vision to classify waste streams in real-time. It uses ResNet50 architectures optimized via the NVIDIA TAO Toolkit for edge deployment, seamlessly merging hardware control with intelligent inference.
- Python
- Deep Learning
- IoT
- Raspberry Pi
- MQTT
PPE Detection System
Status: Production
Thesis: Industrial safety compliance via computer vision.
Industrial safety compliance system built at Stackfusion utilizing Computer Vision and Deep Learning. Features a multi-class YOLOv4 PPE detection pipeline processing live CCTV feeds, accelerated for inference using NVIDIA DeepStream for real-time threat detection and worker safety monitoring.
LLM Verification Workflows
Status: Internal
Thesis: Automated KYC and verification using face detection + LLMs.
At Axis Max Life Insurance, this system automated KYC workflows by fusing Machine Learning face detection with LangChain and ChatGPT APIs. It streamlines policy processing by extracting real-time data, establishing a highly robust verification pipeline designed with Full Stack Development best practices.
- LangChain
- ChatGPT API
- Python
- Face Detection
- Prompt Engineering
Robot M.A.R.V.I.N.
Status: R&D
Thesis: Autonomous navigation robot with LIDAR + ultrasonic sensing.
M.A.R.V.I.N. is an autonomous navigation robot created at the Plaksha YTS program. It integrates LIDAR and ultrasonic sensors for real-time spatial awareness. Capable of complex obstacle avoidance, it sets the foundation for integrating TimeSformer architectures into edge-robotics for temporal environmental understanding.
- Python
- LIDAR
- Ultrasonic Sensors
- ROS
- Embedded Systems
Stock & Vision AI Models
Status: R&D
Thesis: >99% accuracy ML models built on Azure with HPE mentorship.
Developed under the HPE AIYA Scholar program with Nanyang Technological University, these Machine Learning and Deep Learning models achieved over 99% accuracy in Stock Price Prediction and Dog Breed Classification. Built on Microsoft Azure, the Computer Vision workflows represent highly scalable AI solutions.
- Python
- Microsoft Azure
- ML
- TensorFlow
- Data Science