MubarakShikalgar
Machine Learning Engineer
Building intelligent systems at the intersection of Graph Neural Networks, Spatiotemporal Forecasting, and Deep Reinforcement Learning.
Building theFuture of AI
I'm a Machine Learning Engineer and graduate research student with 3+ years of machine learning research and consulting experience. Currently pursuing my Master's at Wilfrid Laurier University in Waterloo, Canada.
My expertise spans Graph Neural Networks,Spatiotemporal Forecasting, NLP, andDeep Reinforcement Learning. I'm committed to delivering ML solutions following sound engineering practices and enabling adoption of ML best practices.
Technical Arsenal
Machine Learning
Maps & Geospatial
ML & Data Libraries
Optimization
Data & Engineering
Research & Methods
Programming Languages
Work Experience
Graduate Research Assistant
SPIN Lab, Wilfrid Laurier University • Waterloo, ON
- ◆Leading LinkETA project: Multi-relational Graph Transformer for link-level travel time prediction
- ◆Built city-scale spatiotemporal ETL pipeline processing 30,000 taxi trips from Porto & San Francisco
- ◆Achieved route ETA MAPE of 18.43% (Porto) and 18.13% (San Francisco) with novel GNN architecture
Graduate Teaching Assistant
Wilfrid Laurier University • Waterloo, ON
- ◆CP421 Data Mining: Guided students on MapReduce, clustering, recommender systems
- ◆CP476 Internet Computing: Mentored students in web development projects
- ◆Evaluated assignments, graded projects, and proctored examinations
IT Consultant
Glazingshopee • Mumbai, India
- ◆Led end-to-end product development achieving 120x faster quotation generation
- ◆Managed 5-person cross-functional team with agile planning methodologies
- ◆Designed automated workflows for IGU configuration, cut optimization, and procurement
Featured Projects
LinkETA: Graph Transformer for ETA
Multi-relational Graph Transformer with relation-specific multi-head attention for link-level travel time and traffic prediction. Achieved 18.43% MAPE on Porto and 18.13% on San Francisco datasets.
DRL Recommendation System
Domain-specific recommender using Deep Q-Network with AutoEncoder, improving MAE by 6.3%
Fake News Detection
Hybrid BiLSTM-TCN with Attention achieving 97% accuracy, presented at IEEE Conference
DTUMOS Dispatch Engine
Real-time ride dispatch optimization using Mixed-Integer Programming with OR-Tools
City-Scale ETL Pipeline
Spatiotemporal data pipeline with GPS map-matching and traffic feature engineering
AutoEncoder DQN-AE
1.2M parameter AutoEncoder model achieving test MSE of 0.0584 for recommendations
Publications & Research
LinkETA: Link-Level Travel Time Estimation Using a Multi-Relational Graph Transformer and Physics-Guided Calibration
Mubarak Shikalgar, Dr. Sukhjit Singh Sehra
Novel multi-relational Graph Transformer architecture for accurate link-level travel time prediction using movement-aware relation graphs.
Fake News Detection Using Hybrid BiLSTM-TCN Model with Attention Mechanism
Mubarak Shikalgar, et al.
Hybrid deep learning model combining BiLSTM and TCN with attention mechanism achieving 97% accuracy on ISOT dataset.
Education & Certifications
🎓Education
Master of Applied Computing
Wilfrid Laurier University
GPA: 10.75/12.0 • ML, Parallel Programming, Algorithm Design
B.Tech in Computer Science
Sanjay Ghodawat University
GPA: 8.29/10.0 • Data Structures, Algorithms, Distributed Computing
🏆Awards & Grants
Graduate Student Funding
$47,950.54 Research Grant • 2024-26
Graduate Studentship
$6,000 Non-Employment Support • June 2025
SGU Best Project Award
Fake News Detection Paper • 2021-22
Let's Work Together
I'm always interested in hearing about new opportunities, collaborations, or just having a chat about ML/AI.
Get in Touch
mubaraks.edu@gmail.com
linkedin.com/in/mubaraks
GitHub
github.com/mubarak-exe
Location
Waterloo, ON, Canada