Aug 2022 – Aug 2023 · 1 yr · Remote · Remote

Jul 2021 – Jun 2024 · 2 yrs 11 mos · Bengaluru, Karnataka, India · On-site

• 1. Developed Customer Cohort identification pipeline using natural language and product knowledge graphs. Using this the teams can find the universe of customers that have a high propensity of being recognized by the cohort. The pipeline also identifies prospective clients who can be introduced to the items falling in the cohort. This is used for targeted notifications, event based category targeting and the personalization team at Blink-, which has increased CTR by 3x and has increased the conversion of users by 2%.
• 2. Owned the data and inference pipeline for Demand Forecast at SKU level at different stores. The project reduced the dump of perishable products to ~5% from 13% while maintaining 97% availability till 9 pm in all stores, pan India.
• a. Reduced time to data prep from ~7.5hrs to ~2hr while also automating it using Airflow DAGs.
• b. Deployed data validation using Great Expectations with slack alerts.
• c. Developed an automated pipeline to train multiple time-series models including Pytorch-TFT, FB- Prophet and tree based regression model using AWS lambda. Due to the automated pipeline the cost of using GPU was reduced from ~3000$ a month to ~1000$.
• d. Built business metrics and sql dashboards, for monitoring impact for stakeholders and for faster debugging.
• 3. Developed scalable ML models for predicting ETAs from dark stores to customers. Model supports a daily load of 10 lakh requests.

Oct 2020 – Jul 2021 · 9 mos

• 1. Developed scalable models for prediction of Demographic, Economical and Human Settlement parameters of different geographies for providing business solutions. Parameters included affluence index , economic potential index, ambient populatoon, traffic flow.
• 2. Developed a graph based analysis model for identifying human Settlement patterns, and distribution of family types across Indian metro cities.
• 3. Developed rural growth index based on identifying patterns in man made structures, Indian census data, government schemes employed in areas and year on year farming yields. Heavily involved in creating data cleaning and scraping pipelines on Airflow.
• 4. Built data quality as a service using Spark to calculate statistical metric using amazon dequee while providing an interface to implement custom metrics with support for data sources like Snowflake, Data Lakes on S3, Oracle. User can also setup custom checks on metrics, done to ensure trust in the data.
• Introducing full observability and orchestration to serverless spark applications Spark-on-K8s
• 5. Built ETL pipelines using Airflow and created a patch for Airflow 10.0.3 to get more granular metrics for generating dashboards and setting up alerts using Prometheus and Grafana.
• 6. Creating data ingestion and exgestion pipeline for S3 to PostGres (with geospatial manipulations) to S3

Feb 2020 – Oct 2020 · 8 mos

Dec 2018 – Feb 2020 · 1 yr 2 mos

Jul 2018 – Sep 2018 · 2 mos

• Developing a Neural Mass Computational Framework to Study Synaptic Mechanisms Underlying Alpha an Theta Rhythms.

Aug 2017 – Dec 2017 · 4 mos

• Implemented Mapper's Algorithm:
• The current version in Python lacked neat visualisation. I worked with Dr Vitaliy on implementing Mapper's algorithm in C++ and improved the visualisation by generating intermediate steps of processing.