Project Portfolio

A comprehensive collection of my software development projects spanning mobile applications, AI integration, bioinformatics tools, and machine learning implementations.

Mobile
Flutter Apps
AI Integration
LLMs & Bots
Bioinformatics
Genomic Tools
Machine Learning
AI Models

Atom – AI-Powered Discord Assistant

Python LLaMA 3 Discord.py
View Project

Project Summary

A custom Discord bot assistant that integrates with a locally hosted LLaMA 3 model via Ollama, designed to help manage daily tasks, reminders, and alerts in a natural and conversational way. Built to provide a persistent, sarcastic Alfred-style assistant with smart task management.

Key Features

  • • LLaMA 3 Integration: Natural, contextual conversation via Ollama
  • • Smart Reminders: Time-parsed and timezone-aware reminder system
  • • Task Management: Add, delete, mark complete via Discord chat
  • • Morning Briefings: Weather, schedule, and music updates
  • • Game Alerts: Notifications when friends start specific games
  • • Persistent Memory: Named memory system with SQLite storage

Technical Stack

Backend: Python, Discord.py
LLM Runtime: LLaMA 3 via Ollama
Database: SQLite
Deployment: Docker, 24/7 uptime

Technical Challenges

LLM Integration

Hosting and querying a local LLaMA model while maintaining relevant, assistant-like responses and personality consistency.

Natural Language Processing

Parsing natural language time input from users and converting to UTC reliably across timezones.

Memory Architecture

Designing a system for persistent, recallable memory and tasks with efficient SQLite storage.

Allele-Specific Primer Generator

Python BioPython Bioinformatics
View Project

Project Summary

A Python-based tool for designing SNP-specific primers that improve genotyping accuracy using mismatch logic and quality control filtering. Automates the entire primer design pipeline from sequence extraction to export-ready results, reducing manual errors in SNP-based PCR experiments.

Key Features

  • • Mismatch-Aware Design: Intentional mismatches at antepenultimate position for allele specificity
  • • Bidirectional Support: Forward and reverse strands with full reverse complement logic
  • • Quality Filtering: primer3-py integration for Tm, GC content, hairpins, and dimers
  • • Primer Ranking: Sorts candidates by quality metrics and PCR suitability
  • • Multiplex Compatibility: Cross-reactivity checks for primer sets
  • • Structured Export: CSV/JSON output with metadata

Impact

Efficiency: Reduces manual primer design effort by 80%
Accuracy: Minimizes amplification errors in SNP-based PCR
Output: Lab-ready primer sets with quality metrics

Technical Implementation

Primer Specificity Optimization

Implemented controlled mismatches at exact locations to increase allele-specific amplification.

Strand-Aware Logic

Designed comprehensive handling for both 5'-to-3' and 3'-to-5' sequences using reverse complements.

Automated Pairing

Built logic to find complementary primers for SNP amplification at specific genomic distances.

Technologies Used

Python BioPython primer3-py pandas

Machine Learning Project Collection

TensorFlow Computer Vision XGBoost
View ML Portfolio

Project Overview

A comprehensive collection of machine learning projects showcasing various algorithms, techniques, and real-world applications across computer vision, regression, and classification tasks. Each project demonstrates different aspects of the ML pipeline from data preprocessing to model deployment.

Featured Projects

  • • German Road Signs Classification: CNN for autonomous driving perception (99% accuracy)
  • • Bike Rental Prediction: Neural network for demand forecasting (R² = 0.96)
  • • Housing Price Prediction: XGBoost with Census data integration
  • • Bank Marketing Analysis: Classification for campaign optimization (91% accuracy)

Technical Skills Demonstrated

Deep Learning: CNNs, Neural Networks
ML Algorithms: XGBoost, Random Forest
Computer Vision: Image Classification
Data Engineering: Feature Engineering

Project Highlights

Computer Vision Excellence

Achieved 100% accuracy on critical STOP sign classification for autonomous driving safety requirements.

Advanced Feature Engineering

Integrated U.S. Census data with housing features to improve price prediction accuracy significantly.

Real-World Applications

Projects address practical problems in transportation, finance, and urban planning with actionable insights.

Tools & Technologies

Python TensorFlow Scikit-learn Pandas XGBoost

Technical Expertise

Mobile Development

Flutter Dart Firebase

AI & ML

TensorFlow LLaMA Scikit-learn XGBoost

Backend & APIs

Python Discord.py SQLite Docker

Bioinformatics

BioPython primer3-py FASTA