Jun 2024 – Jan 2026 · 1 yr 7 mos · Toronto, Ontario, Canada

• Leading the development of Contextual Typeahead, which is an innovative LLM based approach to generating Typeahead suggestions. Building out LLM infrastructure to handle millions of queries per day
• Part of the Text Search team - working on scaling and optimizing our large-scale low-latency Java-based Typeahead ML Ranking Services that receive 1000s requests per sec across multiple Expedia brands (Expedia, Hotels.com, VRBO)
• Designed and implemented multiple improvements to Typeahead that decreased p99 latency by 5x (120 ms to under 25 ms), reduced error rates by 10x, and increased availability from four 9s to five 9s (99.99% to 99.999%).
• Tech stack: Ray, Flyte, Pytorch, XGBoost, Java, Python, Feast, Redis, K6, Recommendation Systems

Sep 2022 – Apr 2024 · 1 yr 7 mos · Toronto, Ontario, Canada

• Built ML systems for Venture Capital

Sep 2020 – Sep 2022 · 2 yrs · Toronto, Ontario, Canada

Dec 2019 – Sep 2020 · 9 mos

• Acquired by Uber in March 2019 for 3.1B USD
• Yoda's current state:
• Serves 10+ million daily realtime predictions
• Hosts 5000+ python based models
• Used by 35+ Data Scientists since the inception of the platform
• Tech Stack: Python, Flask, Airflow, GraphQL, AWS Batch, AutoML, XGBoost, Catboost, Tensorflow, Jenkins, Gunicorn, Gevent, Numpy, Docker, ECS, EC2

Dec 2018 – Dec 2019 · 1 yr

• Yoda now orchestrates the training and serving of numerous models related to ETA, surge pricing, demand/supply, fraud, geospatial predictions deployed across our 100+ cities, with this number skyrocketing as we find more ML use cases. Leading Careem towards a more AI-driven organization

Sep 2017 – Dec 2018 · 1 yr 3 mos

• Founding Engineer for Yoda, our in-house Machine Learning Platform - Responsible for designing & building services to train, deploy and monitor Machine Learning models at scale (AutoML, Traditional ML, Neural Networks, Reinforcement Learning)

Oct 2016 – Sep 2017 · 11 mos · Dubai, United Arab Emirates

• Hacking on Dispatch and Route Optimization algorithms using Golang, Python and Java. Building & Scaling the following products on AWS:
• 1) Order Dispatch service & Driver route optimization (Serverless)
• 2) Knowledge-base based Text address to Geo-coordinates prediction
• 3) Driver telemetry service (Serverless)
• 4) ETA Prediction
• 5) Fraud Detection
• 6) Dashboards for Driver Performance

May 2015 – Jul 2015 · 2 mos · Dubai, United Arab Emirates

• Full-Stack Developer on a Computer Vision/Machine Learning based product for the Xbox Team.

Aug 2014 – Oct 2014 · 2 mos · Greater Hyderabad Area

• Worked on a geo-tagging based product as Android Developer.