Feb 2024 – Jan 2026 · 1 yr 11 mos · Bengaluru, Karnataka, India · Remote

• SQL Pushdown: Built an extensible SQL pushdown framework for BigQuery, Snowflake, and
• Databricks using Apache Calcite, enabling column pruning, filter pushdown, and
• cross-database SQL generation—improving execution speed by 20× compared to
• Spark-based jobs.
• Live Query: Developed Live Query mode in Designer Cloud to enable real-time design-time
• execution on customer CDWs, overcoming prior limitations of partial data access
• and data ingress via async background processing and query cancellation, and
• strengthening partnerships with Databricks, Snowflake, and Google.
• CDF extension: Enhanced the CDF 2.0 framework to support dynamic tool execution through a
• multi-stage reflection and orchestration pipeline, enabling runtime SQL gene
• Iceberg Tables: Integrated read/write support for Iceberg tables via Snowflake Catalog, unlocking
• interoperability for multi-engine lakehouse customers.
• Cloud Native Mode: Built an alternative palette in AAC that executes the workflow in real-time on the customer’s CDW. This was a significant advancement over the previous internal solution, which was limited to a subset of the data. This aided in fortifying our alliances with Databricks and Snowflake.
• Snowpark Containers: Enabled workflow execution within Snowpark (Snowflake’s managed compute) by deploying Alteryx’s engine in-container, eliminating data ingress and reducing latency by 35%.
• Snowflake Private link: Deployed a lightweight service within Snowflake Private Link customer VPCs,
• accessible via control plane port forwarding.

Sep 2022 – Feb 2024 · 1 yr 5 mos · Bengaluru, Karnataka, India · Remote

• Databricks Integration: Integrated Databricks with AAC enabling customers to link their Databricks workspace, read from delta tables, run AAC workflows using Databricks Spark or SQL engine, and write to delta tables. This effort helped onboard 15+ new customers.
• EMR on Shared Data Plane: Having a shared data plane in Alteryx’s VPC for all customers was a stop along the way to having individual private data planes for each Alteryx customer. I worked on setting up AWS EMR in the SDP so that users could launch workflows with only one click.
• Pushdown support for many-to-many cross section: A lot of workflows in Alteryx and their corresponding graphs tend to have many-to-many cross sections. I used graph manipulation techniques to facilitate pushdown in these scenarios. It resulted in the ability to run such complex workflows with data warehouses like Databricks, Snowflake, and Bigquery.
• Reservoir sampling in Spark: Getting a sample of data at any given point in an Alteryx workflow is a critical feature. I implemented Reservoir sampling algorithm using dataframe APIs by leveraging the parallel Spark executors, which resulted in creating samples in excess of 5GB improving from the earlier bottleneck of 100MB.
• Tools Extension: Worked on providing support for newly added tools in Alteryx Designer to run on various engines. This included extending CDF, which is an internal transformation language, implementing UDFs and UDAFs for Spark and SQL based engines, and using Apache Calcite for SQL generation.

Aug 2020 – Aug 2022 · 2 yrs · Mumbai, Maharashtra, India · Remote

• Data Platform: Built a scalable event driven data pipeline using the publisher subscriber model with Kafka event queue system to ingest structured and unstructured (pdf, images) user health data.
• Error Handling and Reprocessing: Designed and built an error handling system for event driven platform which had multiple publishers and subscribers with capability to persist and reprocess failed events.
• Big Data Lake: Developed a real-time ETL analytical pipeline that reads anonymized health OLTP data published into a Kafka stream, then applies transformations written in Scala running and stores it in HDFS using a Spark streaming job. Then I created hive tables over the ingested data, which were used to power dashboards for business use cases.
• Azure Data Lake: Developed a real-time ETL analytical pipeline using Azure services. The OLTP data was published to EventHub, transformed using Pyspark code written in Databricks notebooks, and eventually stored in ADLS. Then I created Databricks tables over the ingested data, which were used to power dashboards.
• Notification system using Firebase: Integrated Firebase Real-time DB that uses SSE technology in background with our application to send instant notifications to the user for any asynchronous activity occurring in the backend.
• FHIR Health Data: Integrated Azure FHIR Service with our product to convert health data in JSON format to FHIR, which is a standard format for electronic healthcare data, and ingested it in a data warehouse that was used to power PowerBI dashboards.
• Logging library: Developed a logging pipeline that collects logs from all services and sends them to the ELK stack using Filebeat. I created a python library tbat can be embedded into Python service to emit structured JSON logs.
• Read Service in Golang: Created a read service in Golang for users health data stored in NoSql databases (Mongo and Cassandra) using goroutines to optimize performance.

Jul 2019 – Dec 2019 · 5 mos · Bengaluru, Karnataka

• Worked on inference and hardware acceleration of pre-trained neural network models on CPU, GPU, and hardware accelerators of android devices using NNAPI (Neural Network API) on an Android Q test device.

Dec 2017 – Jan 2018 · 1 mo · Singapore

• Worked as an Academic Intern under the Global Academic Internship Program at Nanyang Technological University and Hewlett-Packard Enterprise, Asia-Pacific. Learnt 'Big Data Analytics using Neural Networks' and 'Big Data and Hadoop System Administration'. Also did a neural network and AI project on prediction of stock prices depending upon corporate news headlines.

May 2017 – Jul 2017 · 2 mos · Goa, India

• Worked on a Harbour Monitoring Platform called HarMoni. Its main purpose was to monitor pollution, safe management of traffic, and security of the port. Designed a prototype and simulated it in ANSYS software.