Sahil Sanjay Chavan

New York, NY | (917)-254-7789 | [email protected] | www.github.com/Sahil-Chavan | www.linkedin.com/in/sahil-chavan

EDUCATION
Columbia University New York, NY
M.S. in Data Science Expected - Dec 2025
Relevant Courses: Applied Deep Learning, Algorithms for Data Science, Machine Learning in Finance.
Mumbai University Mumbai, India
B.Tech. in Computer Engineering, GPA: 9.29 July 2018 - May 2022
 Won the Best Project Award at a hackathon organized by SDG RAIT, collaborating with a team of 6, with over 40 participating
teams.
 Featured in the Times of India and received a letter of appreciation from the Director at NSS MU for contributions to the NSS Portal.
TECHNICAL SKILLS
Languages & Databases: Python, JavaScript, C, JAVA, PHP, HTML, CSS, Snowflake, MongoDB, SQL
ML/DL Libraries: NumPy, Pandas, Seaborn, Matplotlib, OpenCV, Pillow, NLTK, Scikit-Learn, TensorFlow, Keras, Pytorch, Detectron
Cloud, Analytics & Web Dev: AWS, Domo, Tableau, Django, Flask, NodeJS, AJAX, Bootstrap, Materialize CSS, Postman, GitHub
PROFESSIONAL EXPERIENCE
Teachers Insurance and Annuity Association of America (TIAA) Global Capabilities Mumbai, India
Application Integration and Data Engineer July 2022 – July 2024
 Collaborated with cross-functional teams to orchestrate a comprehensive data warehouse on Snowflake, consolidating over 10+ TB of
data from over 12 sales enablement and marketing platforms, including Salesforce, Hootsuite, ON24, Demandbase, and GA-GMP.
Efficiently managed data across Raw, Trusted, and Refined zones to enhance data accessibility and analytics.
 Developed and automated ETL pipelines for seamless data ingestion and processing using AWS S3, Glue, and Talend, with
scheduling handled by Autosys. This reduced ingestion time by 33%, accelerating access to new data sources.
 Engineered Python and Pyspark scripts to ensure data quality and consistency between source and target systems. The automation
tool was devised to handle multiple data sources and targets, including SQL, Snowflake, Hadoop, and local files, with the capability
to compare data across different connection gateways. This improved data accuracy by 15%, enhancing data reliability.
 Streamlined data pipelines by integrating AWS Sagemaker for machine learning model training and Domo for real-time business
intelligence visualizations, enabling the extraction of actionable insights that enhanced data-driven initiatives across 5 departments.
 Won Hack-A-Cause (Hackathon by TIAA) & 2nd position in the ‘Imitate and Elevate’ by Toastmasters India.
National Service Scheme (NSS) of Mumbai University Mumbai, India
Machine Learning Engineer Intern Mar 2020 – July 2020
 Implemented an end-to-end MERN stack web application, teaming up with stakeholders and volunteers, to automate the selection
process, and visualize volunteer data statistics. Reduced manual effort by 13+ days and aided strategic decision-making in NSS.
 Spearheaded a Volunteer Recommendation module built-in Keras using a Variational Autoencoder-based Collaborative Filtering
algorithm. Optimized 5+ hyperparameters through extensive experimentation, increasing volunteer matching accuracy by 25%.
PROJECTS & RESEARCH
Application of Object Detection & CBIR in the E-commerce Fashion Industry
 As a final year project and thesis for 2 research papers, developed a complete E-commerce solution incorporating CBIR using Image
as a query for buyers and OD and Classifier module for automatic tagging of over 1000+ products benefiting sellers. Performed data
augmentation techniques like geometric, color space, and resolution transformations on scraped data using OpenCV and Pillow.
 Formulated a 256-layer deep CNN in TensorFlow, integrating it into OD models like Faster RCNN and SSD algorithms with transfer
learning. Trained the model on a dual GPU Linux server with 128GB VRAM, achieving a 10% improvement in detection accuracy
compared to previous iterations of models.
 Analysing variational strengths in different model combinations, formulated the ‘Non-linear Maximal Weighted Box Integration’
algorithm to ensemble results from OD multiple models into one superior output in a quadratic weighted function, outperforming
existing models and integration techniques with a 7% increase in overall performance.
Facial Recognition-Based Attendance System (as a Project Lead at SDG RAIT)
 Led a team of 4 to guide a Windows application for facial recognition-based attendance. Utilized MTCNN for face detection and
generated embeddings using ArcFace from 50 real-time images from multiple angles. Employed SoftMax classification and Cosine
distance for facial similarity scoring. One-click new user registration reduced manual input time by over 83%.
PUBLICATIONS & COPYRIGHT
An Ensemble-based Approach for Object Detection Models and its Application in E-commerce Sept 2023
Congress on Smart Computing Technologies (CSCT 2023), Accepted to Publish in Springer
Automated Research Proposal Generation and Approval with Version Control Copyright by Govt. Of India June 2023
FIODC Architecture: The Architecture for Fashion Image Annotation June 2023
Proceedings of International Joint Conference on Advances in Computational Intelligence in Springer, Algorithms for Intelligent Systems