Projects

A collection of my recent work in software development and data science.

Currently in Progress

Skema

An AI-powered SQL schema designer that guides users through data modeling via structured questions, interactive chat, and a visual ERD editor — generating production-ready SQL with smart defaults.

Go
Tailwind
LLMs
Docker
PostgreSQL
Mermaid.js
TypeScript
Progress: 0%
optimyze

AI-powered application that analyzes job descriptions, customizes resumes and cover letters based on user profiles, and generates professionally formatted PDF documents. Features profile management, intelligent job requirement analysis, and document templating system.

Django
LLMs
Langchain
Prompt Engineering
NLP
Progress: 0%

Completed Projects

Hirestack

A scalable, serverless job aggregation platform that processes real-time data from multiple job boards and provides intelligent job recommendations using machine learning.

Python
Django
Next.js
AWS
DynamoDB
Machine Learning

Type

Full-Stack Web Application

Key Achievements

  • Reduced infrastructure costs by 40% through serverless architecture
  • Improved job matching accuracy significantly by using ML algorithms

Details

  • Implemented serverless architecture using AWS Lambda and API Gateway
  • Designed and optimized DynamoDB schema for efficient querying
  • Developed machine learning models for personalized job recommendations
  • Integrated with multiple job board APIs for real-time data aggregation
  • Implemented user authentication and profile management features
RetailPulse

An IoT-based People Counter application using computer vision for accurate real-time customer tracking in retail environments, providing actionable insights for store optimization.

Python
OpenCV
FastAPI
Docker
PyTorch
IoT

Type

Computer Vision & IoT Application

Key Achievements

  • Achieved 95% accuracy in people counting
  • Reduced manual counting labor
  • Provided actionable insights leading to 15% increase in store efficiency

Details

  • Developed computer vision algorithms for accurate people counting
  • Implemented real-time data processing pipeline using FastAPI
  • Designed and built responsive dashboard for visualizing foot traffic data
  • Integrated with IoT devices for seamless data collection
  • Containerized application using Docker for easy deployment
Mood of Music

A web application that generates personalized Spotify playlists from environment images using AI for mood detection and song recommendations

Python
Flask
PyTorch
Spotify API
OpenAI API
PostgreSQL
Docker
Render.com

Type

AI/ML Web Application

Key Achievements

  • Created dual recommendation system using both image analysis and user's top artists
  • Implemented containerized deployment with Docker supporting both PostgreSQL and MySQL
  • Developed responsive web interface for seamless playlist embedding and playback

Details

  • Implemented image analysis using PyTorch to extract emotional parameters from uploaded images
  • Integrated OpenAI API for image description and playlist generation based on detected moods
  • Developed Spotify API integration for playlist creation and management
  • Built user authentication system with Spotify OAuth
  • Implemented database storage for user accounts and playlist details using PostgreSQL
CryoCast

Long-Term Prediction Framework - An AI forecasting system that 'freezes' current trends and extrapolates them into the future with comprehensive uncertainty quantification.

Python
TensorFlow
PyTorch
Pandas
Matplotlib
Streamlit

Type

AI/ML Application

Key Achievements

  • Demonstrated improved forecast consistency compared to standard time-series models
  • Created visualization tools that effectively communicate prediction uncertainty
  • Implemented a modular architecture that supports multiple forecasting approaches

Details

  • Architected a machine learning framework for long-range forecasting that captures trend dynamics and quantifies prediction uncertainty across multiple time horizons
  • Implemented ensemble methods combining statistical time-series models with deep learning approaches to improve reliability of extended predictions
  • Designed a visualization system that transparently communicates confidence intervals and probability distributions as forecasts extend further into the future
  • Incorporated multiple data sources to enhance prediction accuracy and provide contextual insights