I teach machines
to think.
Bridging academic research and production engineering. A 4x IEEE published author and MS Data Science candidate at UB, building interpretable ML pipelines deployed in real environments.
Core Stack
My ML Process
From raw data to production-ready models, here's how I build end-to-end machine learning systems.
About Me
Data Scientist
AI/ML Specialist
Full-Stack Developer
Python & Web Tech
Published Researcher
4 IEEE Publications
I'm obsessed with building intelligent systems—whether it's predicting graduate admissions with 85% accuracy or detecting wildlife in complex environments. I specialize in turning messy, real-world data into AI systems that actually hold up in practice.
Currently pursuing my MS in Data Science at University at Buffalo, where I work with Dr. David Doermann on admission prediction systems. I've published 4 IEEE papers on optimization algorithms and explainable AI, and I'm always looking for the next interesting problem to solve.
When I'm not training models...
Work Experience
Where I've applied data science to solve real problems.
- Developed an end-to-end admissions ML pipeline processing 3,000 applications per cycle, reducing approximately 300-400 manual review hours and admissions processing time by around 10% for a pilot CSE department using AHP-based interpretable feature engineering.
- Standardized and clustered 100+ academic major titles, reducing preprocessing effort by 20% for admissions data preparation workflows still in testing using TF-IDF, K-Means, and fuzzy matching pipelines.
- Trained Random Forest and GPU-accelerated PyTorch models on approximately 20,000 applications, improving minority-class recall and F1-score by 15% during model validation using Focal Loss to address class imbalance.
- Achieved 85.1% accuracy, 95.9% recall, and 90.7% F1-score across roughly 20,000 applications, automating approximately 20% of rejection reviews while improving fairness via feature removal using LIME-based model explanations.
- Optimized Django application performance, improving response times by 30% and reducing load times for approximately 1,000 users in data-heavy workflows through query optimization and caching strategies.
- Integrated REST APIs for internal and external systems, reducing data latency by around 10% and improving load performance for large datasets in production applications using standardized API-based communication mechanisms.
- Implemented code review checklists, improving review throughput by 50% and reducing post-deployment issues across a five-developer engineering team by enforcing clean-code and review standards.
Education
Building a strong foundation in data science and computer science.
MS in Data Science
University at Buffalo, SUNY
Buffalo, NY, USA
Bachelor of Technology in Computer Science
MIT World Peace University
Pune, India
Featured Projects
Real-world applications of machine learning, from research to production systems.
Publications & Papers
Peer-reviewed research contributions in machine learning and optimization.
Deep Learning for Exoplanet Exploration
ICDAI 2025 - Springer Nature
ANN and Gradient Boosting pipelines achieving 88.3% detection precision and 91.06% habitability prediction on NASA data.
Capacitated VRP using Ant Colony Optimization
IEEE 2025
Capacity-aware logistics routing reducing travel distance by ~2,000 km and cost by ~20,000 units using ACO.
YOLOv5/YOLOv8 for Bird Species Identification
IEEE 2025
Comparative analysis of object detection models for wildlife identification in complex environments.
Fake Profile Detection Using Machine Learning
IEEE 2024
ML-based approach for identifying fake social media profiles with high accuracy classification.
Interactive Tableau Dashboards
Explore my interactive Tableau dashboards showcasing data analysis and insights.
Professional Learning Journey
Certifications and coursework demonstrating hands-on learning and validated credentials.
What People Say
Verified recommendations from colleagues and mentors on LinkedIn.
View All on LinkedIn"I worked with Deven during his internship at Markytics, where he consistently demonstrated strong data science and analytics capabilities. He was particularly effective in using Python and SQL to analyze data, build models, and support data-driven decision-making across projects."
"Deven showed a solid ability to translate business requirements into structured analytical solutions and automate repetitive workflows, improving efficiency and reliability. His approach to modeling and analysis was thoughtful and well-executed, with a clear focus on producing actionable results rather than theoretical outcomes."
"Beyond his technical skills, Deven worked very well within the team. He communicated clearly, collaborated effectively with both technical and non-technical stakeholders, and took ownership of tasks while remaining receptive to feedback."
Received January 2026
"I had the pleasure of supervising Deven Shah during our collaborative research on the Capacitated Vehicle Routing Problem (CVRP) project, which later culminated in a successful IEEE publication. Throughout this period, Deven consistently demonstrated exceptional analytical depth, strong modeling skills, and a remarkable ability to translate theoretical concepts into practical, data-driven solutions."
"His work on developing and fine-tuning optimization algorithms showcased not only his technical proficiency in Python, machine learning, and heuristic modeling, but also his keen understanding of data integrity and real-world application constraints."
"Beyond his technical strengths, what truly sets Deven apart is his collaborative approach. He is a thoughtful team player who elevates discussions with critical insights while remaining open to diverse perspectives."
Received January 2026
Let's Connect
Got an interesting ML problem? Building something cool? Let's chat!