Gabriel Kinshuk
Machine Learning Engineering & Data Science
I build rigorous, scalable data pipelines and machine learning systems for high-stakes research and business environments. By combining statistical modeling with mathematical optimization, I transform complex data into actionable insights that support strategic decision-making.
My work spans automating research workflows, developing computer vision infrastructure, and applying optimization to solve supply chain challenges. Across every project, I prioritize reproducibility, engineering rigor, and measurable impact.
Professional Experience
Computer Vision Capstone Project
Industry Partner: Google
In Progress
- →Conducting a literature review on action recognition techniques in computer vision and analyzing EHS regulations across different regions.
- →Performing data augmentation on training datasets to improve model robustness and support future experimentation.
Project Goals / Deliverables
- →Prototype a computer vision pipeline to detect EHS-related incidents and automate structured data extraction for reporting.
- →Develop models for action recognition and temporal sequence analysis to identify potential causes and risk factors.
- →Establish modular ML workflows and infrastructure to support iterative experimentation and reproducibility.
Graduate Researcher
Biomedical Engineering Research
Designed and implemented automated data orchestration systems for complex research datasets. Built Python-based pipelines emphasizing modularity, reproducibility, and automation.
- →Constructed a data pipeline in MATLAB and Python utilizing machine learning and statistical modeling to determine module compatibility in hybrid transcription regulators.
- →Published two review papers and submitted an experimental paper focusing on predictive modeling for genetic circuits, funded by a government grant.
- →Composed a 70-page research paper and technical documentation in LaTeX, including self-designed figures to illustrate complex biological and statistical models.
- →Automated data processing workflows and implemented contemporary analysis techniques, resulting in a several-fold increase in data throughput.
Education
Master of Science in Business Analytics
University of Texas at Austin
In Progress
→Advanced Machine Learning & Predictive Modeling
→Statistical Analysis & Data Mining
→Business Intelligence & Analytics
Master of Science in Biomedical Engineering
University of North Texas
Completed 2025
→Computational Research Methods
→Data Pipeline Development
→Research Automation & Infrastructure
Projects
TileShift
Puzzle game with procedural level generation using BFS state-space search and solvability verification. Features flood-fill algorithms, comprehensive testing, and MVC architecture.
NYC Transit Analysis
A Tournament of Models for NYC MTA ridership forecasting. Evaluates ARIMA and Bayesian AR(7) to predict subway, bus, and bridge demand for transit supply chain optimization and capacity planning.
Applied Optimization for Business
Collection of mathematical optimization projects focusing on Linear Programming (LP), Nonlinear Programming (NLP), and integer-based decision modeling.
Multimodal Price Prediction
A multimodal machine learning pipeline to predict ASOS apparel prices by fusing tabular metadata, NLP (BERT), and Computer Vision (ConvNeXt) in a Late-Fusion architecture to capture brand prestige and aesthetic value.
Protein Parser
Bioinformatics tool with a GUI built on top of BioPython for parsing and analyzing protein sequence data.
Austin Restaurant Whitespace
Market analysis identifying underserved Austin restaurant segments by transforming unstructured images into semantic labels. Features a computer vision-to-NLP pipeline using TF-IDF and Cosine Similarity to evaluate aesthetic consistency and market positioning.