Data & Machine Learning Professional with 6+ years of experience spanning Operations → Business Optimization → AI/ML Engineering. I bridge the gap between business understanding and cutting-edge AI technology, having delivered measurable impact including 25% revenue increases and $500K+ cost savings.
📈 Career Evolution: Started in operations, advanced through business optimization and analytics, now building production ML systems with expertise from University of Toronto and Stanford ML programs.
🔬 Research Interests: Multimodal AI, Computer Vision, NLP, and Reinforcement Learning
🛠️ Current Focus: Production ML systems, MLOps, and real-time AI applications
💼 Proven Impact: 6+ years driving data-driven decisions across multiple industries
🌍 Open to: Data Science, ML Engineering roles, research partnerships, and innovative AI projects
FrankenBERT: When AI Specialists Collide (Model Merging - Part of Cohere Research Scholar Application)
- Abstract: This research investigates what happens when two AI specialists are merged into a single model. By training separate "Poet" (sentiment analysis) and "Scientist" (news classification) models, then combining them through parameter averaging, we discovered catastrophic interference - the merged model performs poorly on both tasks despite each specialist achieving >90% accuracy individually.
- Key Finding: Simple model merging destroys specialist expertise rather than combining it, with performance dropping by 36-83% across tasks.
- Live Demo: HuggingFace Space
- Tech Stack: Whisper, mBART-50, T5, PyTorch
- Impact: 95% accuracy on noise-resilient speech recognition
- Live Demo: Customer Analytics Platform
- Business Impact: 15-20% churn reduction, 10-15% revenue increase projections
- Features: Real-time customer scoring, interactive UI deployment
- Achievement: 99.5% recall accuracy (5 misclassifications out of 3,817)
- Tech Stack: Random Forest, XGBoost, Feature Engineering
- Application: Content moderation and recommendation systems
- Focus Areas: Computer Vision, NLP, Transformer Architectures
- Frameworks: PyTorch, TensorFlow, Hugging Face Transformers
- Business Impact: $500K stockout prevention, 25% cost reduction
- Tools: Python, SQL, Tableau, Advanced Analytics
🏫 Formal Education
- 🏛️ University of Toronto - Data Science & Machine Learning Certification
- 🏛️ Stanford University - Machine Learning Specialization
- 🏛️ University of Windsor - Master of Applied Science - Electrical Engineering
- 🏛️ Anna University - Bachelor of Engineering - Electronics and Communication Engineering
🏅 Professional Certifications
- 🔬 IBM - AI Developer Certification
- 🤖 NVIDIA - AI Operations & Infrastructure Fundamentals
- 📊 Wolfram Research - ML Statistical Foundations Professional Certificate
- 📈 Google - Advanced Data Analytics Professional Certificate
- 🏛️ University of Pennsylvania - AI, ML Essentials & Statistics
- 🐍 OpenEDG Python Institute - Programming with Python Professional
- 🏛️ Ludwig Maximilian University Munich - Competitive Strategy & Organization Design
- ☁️ AWS - Cloud Technical Essentials
- 🐧 Canonical - Linux Professional Certification
🎯 Model Performance Excellence
- Video Classification: 99.5% recall accuracy on TikTok content moderation
- Speech Recognition: 95% accuracy on challenging multilingual audio
- Customer Segmentation: 15-20% projected churn reduction
💼 Business Impact
- Revenue Growth: 25% increase through predictive analytics at Mama Earth Organics
- Cost Optimization: $500K stockout prevention + 20% waste reduction at Whole Foods
- Operational Efficiency: 15 hours weekly saved through automated reporting systems
🚀 Open Source Contributions
- HuggingFace Deployments: Multiple live ML applications serving real users
- Production ML Systems: End-to-end pipelines from data ingestion to model deployment
- Research Focus: Active exploration in Deep Learning and Reinforcement Learning
🎯 Contribution Stats
