Software Engineer, Google
Jul 2022 - Present, Waterloo, ON
- Working on the GCloud - cost optimization team
ML Engineering Intern, Wish
Sep 2021 - December 2021, San Francisco CA, (Remote)
- Built a k8s service to predict the lifetime value of ~200M users on a daily basis to be used for return on ad-spend prediction, financial forecasting and A/B test analysis for better user targeting
- Onboarded and supported 1 new intern, and 3 full time data scientists and engineers to the team
- Setup core ETL infrastructure to enable Spark/Airflow to work with new in-house data lake
Software Engineer - Research Assistant, ECE, University of Waterloo
May 2021 - Aug 2021, Waterloo, ON
- Worked under Prof. Derek Wright to build a internal platform using Postgres, Node and React for managing the ECE departments’ CEAB requirements
Data Scientist Intern, Wish
Jan 2021 - Apr 2021, San Francisco CA (Remote)
- Built a 94% accurate model to predict the probability of the user refunding an item within 90 days
- Reduced refund abuse by 30% by developing a single scoring mechanism for each user to accurately reflect the abuse by them and their linked accounts against Wish refund policies
- Improved revenue by 2% by using XGBoost to conduct explainable user segmentation to isolate clusters of users that purchase upto 10x more from user generated content
Partner, Technology Lead, Sage Co
May 2018 - Feb 2021, Waterloo, ON
- Designing the online presence and building websites for SMBs in North America (portfolio to date includes community websites, local newspapers, law firms, and small print shops)
- Responsible for managing the technical direction of the company, and leading the charge on all things technical
- Services include branding, web design & development, custom CMS, website audits, Shopify stores, and SEO optimization
Software Engineering Intern, Wish
May 2020 - Sept 2020, San Francisco CA (Remote)
- Improved transaction success by 3% for an annualized improvement in GMV by $15M by setting up remove item retry mechanism for insufficient fund transactions
- Improved traceability of fraudulent actors by marking user sessions with a multi-factor authentication session key to log high risk events
- Increased average customer satisfaction by 3% by designing a user flow to allow users to change their payment method after placing an order
- Setup infrastructure using Pytest fixtures to enable improved testing coverage for key code paths
- Reduced fraud through post-purchase address change by incorporating address verification flows
Software Engineering Intern, KitchenMate
Sept 2019 - Dec 2019, Toronto, ON
- Improved legibility and reduced frame jitter by 50% by adding support for on device text and custom animation rendering for new display on proprietary cooker using Python and Pillow
- Built a top level overview dashboard with all internal high level KPIs in Metabase/PostgresQL to allow for effective decision making process on our corporate customers in a single glance
- Increased customer food satisfaction by 30% by devising a data driven approach to menu design after analysing dish reviews to isolate metrics that better define the popularity of a dish
- Developed over 70 dashboards in Postgres to measure KPIs across the board to promote data driven decision making at all levels of the company
Data Engineering Intern, DraperAI
Jan 2019 - Apr 2019, Waterloo, ON
- Reduced customer conversion time by 15% by building a tool to entice new customers
- Improved bid accuracy by 30% by analysing suggesting changes to core bidding algorithm
- Found 3 high priority production issues affecting 10% of our marquee customers
- Built an internal tool to detect and alert system downtime using SQL and Metabase
ML Engineer - Research Assistant, Machine Intelligence Lab
Sept 2018 - Dec 2018, Waterloo, ON
- Built a NLP chatbot in Python and Dialogflow for grocery chain Loblaws under Prof Fakhri Karray to allow customers to ask contextual questions while in store
- Ported over existing from slower excel sheet frameworks to PostGresQL to improve net performance by 30%
- Developed a low latency API endpoint in Python from scratch to serve Loblaws customers
Data Scientist // Full Stack Developer Intern, LCBO|next
May 2018 - Sept 2018, Kitchener, ON
- Wrote complete CRUD web apps and REST APIs from scratch with less than 0.1% downtime to serve data to multiple internal applications to serve LCBOs customers
- Managed and maintained the sole Kubernetes cluster running an Elasticsearch instance
- Implemented an 85% accurate sales forecasting model from scratch to anticipate uptick in sales of products after incorporating the weather forecast
- Visualised and analysed sales trends at top retail locations to classify products into priority based classes to support LCBO’s retail and supply chain divisions
Core Perception Member, WATonomous
Jan 2018 - May 2018, Waterloo, ON
- Core Member of the perception team as part of GM’s AutoDrive Challenge
- Implemented a 5% more effcient model to identify lane markings in OpenCV
- Improved existing traffic sign detection computer vision models in Tensorflow by 15% using hyperparameter tuning and regularisation
Web Lead, Waterloop
Sept 2017 - Mar 2018, Waterloo, ON
- Redesigned the entire new website as part of revamped branding in Fall 2017 with the help of Embedded JS templating on a Node.js server
- Minimized code duplication on the new website using a template engine that allowed for 30% page load boost while making it easier to maintain
- Improved page load time by 20% by creating responsive vector images in D3.js instead of using heavier and less efficient frameworks
- Implemented an efficient static string method to improve memory management as part of the embedded systems team
Software Engineer Intern, FINO Bank
Summer 2017, Mumbai, MH
- Improved efficiency of backend script by 300 times by using hash tables and used this to implement an interactive dashboard to visualise revenue of top merchants and their sales breakdown
- Conceptualised, designed and implemented a dashboard to track the performance of new merchants month on month to predict merchants who had a better chance to increase the revenue
- Used d3js for the visualisation and Python for the back end.