Jan 2024 – Present · 2 yrs 2 mos

• Uber AI Platform - Building & Scaling Michelangelo ( https://eng.uber.com/michelangelo-machine-learning-platform )
• 1. Working on Control plane for Batch Jobs ( Spark & Ray ) for ML Workloads - Distributed Training Infrastructure for Michelangelo
• 2. Michelangelo Pipeline Debugging agent - Agent that figures out the root cause for failiures in ML Platforms training, feature prep, evaluation, retraining pipelines etc... using SOTA LLM's.
• 3. Architecting and managing the core API server and controllers for Uber's ML Platform, enabling scalable, reliable lifecycle and resource management on Kubernetes.
• 4. Lead, design & implement Responsible AI Framework for Uber's AI Platform - Using AI to make AI responsible at Uber.
• 5. Machine Learning Pipeline Monitoring

Jun 2022 – Mar 2024 · 1 yr 9 mos

• Uber AI Platform - Building & Scaling Michelangelo ( https://eng.uber.com/michelangelo-machine-learning-platform )
• 1. Launched Generative AI focussed features in Uber's AI Platform - Monitoring Quality and evaluation of LLM's for various usecases, Model Garden for Generative AI Models
• 2. Machine Learning Model Quality, Observability and Monitoring - Introduced Metric Framework for standardized ML Quality measurement, realtime/offline analytics and monitoring via alerts. Built K8s and spark based pipelines for Data ingestion.
• 3. ML Workflows ( Retrain Pipelines ) using K8s based pipelines.

Feb 2021 – May 2022 · 1 yr 3 mos

• Uber Eats fares Platform - Backend Engineer, also contributed to ML and Data at Uber. We decide how the fares are computed in a distributed environment for an Uber Eats order.
• 1. Business model Changes - Worked on critical business model changes in countries like Canada, Mexico to increase Uber's profitability in these markets.
• 2. Time series Forecasting & Anomaly Detection - Backtested a POC and built a time series based anomaly detection prototype ( using statistical + deep learning models ) to detect outages by having alerts over business metrics in different timezones where Uber Eats operates.
• 3. Scalable and configuration driven Tax profile entry system for Uber Eats customers. Built a system in Uber eats to allow customers to optionally provide their tax id in eats checkout & claim tax credits. This feature spanned more than 10 countries where each tax profile had very unique properties.
• 4. Merchant Recommendation Engine - Built a prototype for hierarchical geospatial index ( h3 ) based recommendation engine to provide insightful recommendations to the merchants.

Aug 2019 – Jan 2021 · 1 yr 5 mos

• Uber Eats Payment Integration Team - Full-Stack Engineer( Backend & Android ), Built and extended API's ( Golang, Java, Python ) under Uber's payment ecosystem. Integrated multiple payment methods in Eats - Bancontact, UPI & Edenred-Brazil.
• 1. Uber Pay - UberPay allows payment providers to integrate with Uber’s payment ecosystem by implementing a set of published API specs.
• 2. Migrations
• a. Node.js Edge layer to Golang microservice based presentation layer.
• b. Android Dagger to Uber Motif Dependency Injection framework migration
• 3. Built an ultrasound based OTP exchange prototype in a Device 2 Device communication Hackathon. Rank - 1 prize across Uber ( Global ), with more than 200 teams. The concerned team then went on to build the feature for production.