Windfall is the platform for modern wealth intelligence. As a part of the platform team, worked

on various initiatives to improve data validation, monitoring, and ingestion of customer data.

• Implemented a match-validation job to flag potential bad records based on various heuristics, using Apache Spark. Successfully scaled to process over 12 million records and built pre and post processing jobs in Java to handle input table creation, streaming results to BigQuery and Postgres. This implementation improved the company's data quality, reducing customer churn rate by 18% and also saving the company approximately $150,000.

• Implemented alerting functionality in Windfall’s custom event driven orchestration framework which runs data processing jobs for customers, using Kotlin thus reducing the number of support tickets created by customers.

• Implemented a new parser to parse customer CSV files to JSONL (json-lines) using Kotlin. The CSV files were of various sizes ranging from a few megabytes to multiple gigabytes. Improved CSV parsing time using the Apache Commons CSV library and decreased the turnaround time for processing customer data.

Contributed significantly to the Ads Monetization Platform’s engineering team, implementing core features to improve scalability, availability, and performance. The platform serves real-time ads to 300M+ MAUs across ABM’s game portfolio, processing 55,000+ requests/sec, and generating $400M+ ARR.

• Developed a feature in the campaign manager in Java to group ad placements by name, enabling accurate tracking of ad performance by game and audience.

• Built a publish-subscribe system using Google Pub/Sub in Java to enable real-time updates of ads configuration data from the campaign manager to the mediation service, reducing update latency from 2 hours to 30 seconds via MySQL-based distributed locking.

• Designed and implemented a scheduler job to upload incremental experiment view changes to Google Cloud Storage, improving update frequency from 24 hours to 15 minutes, enhancing downstream data consumption.

• Improved scalability and robustness by removing session affinity in the mediation service, enabling ad requests to be processed by any pod. Utilized Redis for user data storage and distributed locking, preventing overload on game servers.

• Separated experiment and whitelist view generation jobs, allowing independent uploads to Google Cloud Storage, streamlining A/B testing and audience management.

• Spearheaded the creation of a Java library to externalize application properties, enabling updates without redeploying services, and conceptualized and established a central config store to decouple configuration from code.

• Developed a JavaScript generation framework to dynamically override ad behavior in games using an in-house query language (AQL).

• Refactored legacy Java code to optimize memory usage and streamline event delivery to the event service, improving system performance.

Mapbox's Search POI (Points of Interest) team was responsible for creating data pipelines to ingest maps data from various sources and load the maps data into the data warehouse. The data warehouse contains all of the information necessary to power the search, and Mapbox generates about $100 million in ARR.

● Spearheaded the development of the Foursquare v3 data ingestion pipeline, consuming roughly 600GB of maps

data into the data warehouse using Python/PySpark, Apache Airflow, and AWS. This pipeline boosted customer engagement in search results by 21%.

● Designed and implemented logic to deduplicate and conflate records of map data from various vendors, including Foursquare, Safegraph, and OSM, resulting in a 12% decrease in errors in customer search results.

● Delivered high-quality maps data to downstream teams and created internal tools to streamline pipeline

operations, significantly improving efficiency.

● Conducted interviews for 10+ candidates, contributing to team growth and talent acquisition.

Cisco's Ultra Cloud Core is a solution that supports telecom 3GPP standards for 5G. The 3GPP contains many standards and specifications. The standards contain a definition of the Policy Control Function (PCF) network function. PCF is responsible for enforcing policy control and providing the appropriate Quality of Service (QoS). It is one of the core network functions for 5G. I worked as a Software Engineer on the PCF team. PCF is a cloud-native solution.

● Developed key features for the PCF policy engine, enhancing service discovery, load balancing, and request processing to improve system efficiency and scalability.

● Resolved a critical scalability issue in production by identifying and fixing an OutOfMemory error caused by 100+concurrent connections to the PCF REST endpoints, ensuring system stability.

● Designed and implemented a microservice in Golang to generate Yang, Render, and ConfigMap files for

customers using Kubernetes, scaling to 35+ columns and 4000+ rows. This microservice eliminated human error in complex 5G configuration deployments.

● Developed a Vue.js UI feature for exporting custom reference data stored in MongoDB, improving data

accessibility for users.

● Engineered a critical feature in a multi-threaded network-cutter tool, distinguishing IPV4/V6 connections and identifying connection drops between server-side components (PCRF) and the User Data Cache (UDC). This feature enabled up to 50 peers to be handled simultaneously for improved troubleshooting.

● Interviewed 8+ candidates, contributing to team growth and technical hiring efforts.