Building infrastructure & tools for Ads Reliability. Leveraging machine learning and AI for revenue forecasting, detecting and diagnosing reliability events and estimating impact.

Building data and ML platform, ML applications and realtime ML services to bring the best photos and rich data experience to Apple Maps places.

Deep experience in designing and building end to end Maps features. Cross-functional collaboration with business, product and engineering teams. Adept at roadmap planning, facilitating design discussions, and providing technical direction and mentorship.

Built an MLOps platform from scratch to reduce operational overhead and facilitate faster iterations in ML lifecycle. Involved in gathering user requirements, translating to a technical design, making technology choices and implementation. Design based on ease of use with abstracted interfaces that allow specifying an ML workload in JSON style declarative schema without needing any knowledge of compute infra. Hacked a POC within a day to provide confidence on the design.

Built a coarse parallel compute platform enabling seamless transition from laptop to cloud for data science team enabling them to parallelize their workloads on Kubernetes clusters

Developing predictive models for components of deal recommendation system.

Building data pipelines(batch & streaming) for feature engineering using Spark(Scala API).

Analyse data and propose prospective features likely to improve accuracy of ranking model.

Perform periodic on-call duties like handling data pipeline issues and code deployments.

Inventor on multiple patents filed by Groupon

Teaching Assistant in Undergraduate course on Probability and Statistics.

Conducted discussion sessions on topics such as probability distributions and bounds, random variable statistics, statistical tests etc.

Trained and developed a predictive model for the deal recommendation system used by Groupon, specifically, a model to predict the repeat purchase behavior of users for making a choice between explore/exploit treatment for the user.

Wrote Spark jobs to collect and process over 1Tb data.

Efficiently implemented the machine learned model online to make real-time recommendations.

Successful A/B tests showed highly significant improvement in recommendations.