Sanghamesh Shiddalingesh Vastrad

Sanghamesh Shiddalingesh Vastrad

Data Scientist

University of Toronto

Biography

Hi! I’m a Data Scientist and an Applied ML Researcher. As much as I love building unique, state-of-the-art, deep learning solutions, I also love the data science behind experimentation and visualization.

Experienced in all aspects of the Machine Learning workflow from feature engineering to model explainability, I’m currently a Machine Learning - Applied Research Intern at ICICI Bank working on credit underwriting models and fraud detection. I’m also an MSc. in Applied Computing candidate specializing in Data Science.

Interests

  • Deep Learning
  • Data Visualization
  • Self Supervised Learning
  • Randomized Control Trials (A/B Testing)
  • Data-driven Product Management

Education

  • MSc in Applied Computing (Data Science Concentration), 2020

    University of Toronto

  • BEng in Computer Science, 2018

    Sri Jayachamarajendra College of Engineering

Skills

Python for Data Science & ML

NumPy, Scikit-learn, Pandas, xgboost, lightgbm, imblearn, StatsModels

R for Data Science & ML

dyplr (dplyr, ggplot2, broom)

Data Visualization

Plotly, matplotlib, seaborn, ipywidgets, Voila, streamlit

Deep Learning

Tensorflow, Keras

Big Data Engineering

Hadoop, BigQuery, Spark, SQL

Application Development

Java, C#, C, JavaScript

Version Control

Git, Github, Bitbucket

Experience

 
 
 
 
 

Machine Learning Applied Reseacher Intern

ICICI Bank Canada

May 2020 – Present Toronto
  • Sole member of the Data Science and Analytics Group at ICICI Bank Canada with research approved and partly funded by Mitacs Accelerate. Trusted with designing and building a ‘Zero Credit Touch’ (ZCT) system with projected impact on 20 million+ customers.
  • Improved credit underwriting gini coefficient (2*AUC - 1) by 12% and decile wise gini by 6% by developing an ensemble model consisting of XGBoost, LightGBM and a neural network (with LSTMs and 1D Convolutions).
  • Cut down data dimensionality by 55% from 400 to 185 features using Shapley model explanations and manifold learning techniques like Locally Linear Embeddings and t-SNE.
  • Optimized AUC using synthetic data generation using a combination of SMOTE and CTGAN (Conditional GAN for Tabular Data) after researching best-suited techniques for handling imbalanced datasets.
  • Used Latent Dirichlet allocation (LDA) for customer segmentation to identify clusters where the model does not perform well, that translated to an increase in model’s decile wise gini by 11% when the poor performing cluster is removed.
 
 
 
 
 

Software Developer

Western Digital

Jul 2018 – Jul 2019 Bangalore, India
  • Delivered key features for managing, verifying, and visualizing iNAND firmware algorithms using C# and XAML in an agile environment meeting over 90% deadlines.
  • Spearheaded prototyping for ‘image-save’ capability for SSDs of 2TB+ capacity using Spark and Kafka to make data processing 400% faster.
  • Led data visualization tasks using Live Charts for Windows Presentation Foundation (WPF).
 
 
 
 
 

Software Engineering Intern

Western Digital

Jan 2018 – Jun 2018 Bangalore, India
  • Reduced version control testing time from 72 to 8 hours by designing and implementing an automation framework using Python.
  • Presented over 20 data analysis results and reports of multiple HDD performance benchmarking programs.

Accomplish­ments

Certified Data Scientist

Workera Jul 2020
See certificate

Advanced SQL: BigQuery

Kaggle Jul 2020
See certificate

Deep Learning: Sequence Models

Coursera Jun 2020
See certificate

Machine Learning Explainability

Kaggle Feb 2020
See certificate

Runner Up : Code for Good hackathon

J.P. Morgan Chase & Co. Jul 2017

Huawei Technologies Scholarship

Huawei May 2016 – May 2018
Two-time recipient of Huawei Technologies scholarship (worth 50,000 rupees) for being one of the top 10 (out of 800+) students in the college.

Rank 2

Sri Jayachamarajendra College of Engineering Jul 2014 – May 2018
Awarded multiple scholarships including Ministry of Human Resources Development India Scholarship and Shashikiran Kadaba endowment award

Recent Posts

Projects

Averting Algorithm Aversion Through Explainability

Conducted an online lab (A/B testing) experiment through Qualtrics Survey Platform and Amazon MTurk to determine the effect of model explainability on machine learning model aversion.

Recent & Upcoming Talks

Recent Publications

Quickly discover relevant content by filtering publications.

Popular Topics

A/B Testing

Contact

  • Bahen Centre for Information Technology, 40 St George St, Toronto, Ontario