Pay Platform team

Part of a team responsible for all money movement (payins & payouts) within Lyft processing billions of dollars worth money movement annually

Refund reliability

• Engineering driven project to improve refund reliability, visibility, and ability to retry failed refunds

• Refund failure rate fell rom a peak ~16% in May 23 to ~0.3% in October 23.

• Resolved a backlog of thousands of refund failures affecting dev productivity for the team

• Saved ~2 hours of time for support during incidents

• Delivered tool that visualize refund failure reasons to support teams

• Saved ~100k/month worth of money due to customer filing disputes on failed refund

Line item management

• Created a tool within Lyft to help reduce complexity in approving and creating line item when launching new Lyft products

• Process resulted in slashing the time to approve + create line items from 4 weeks to a couple of hours boosting dev productivity and launch time across Lyft

AWS Aurora Postgres Upgrade

https://eng.lyft.com/postgres-aurora-db-major-version-upgrade-with-minimal-downtime-4e26178f07a0

Deployed RDSProxy for critical services within Lyft

https://eng.lyft.com/beyond-query-optimization-aurora-postgres-connection-pooling-with-sqlalchemy-rdsproxy-200db7f562d7

Resolved memory leaks tied to Python + Gunicorn for number of teams within Lyft

Working on building the next generation Customer Data Platform

Some of the technologies I got my hands on: Java, Spring, Hibernate, Docker, Kubernetes, Spinnaker, Helm, Terraform, gRPC, OpenAPI

• Technology: Python, Django, Tastypie, Java, Jenkins, Docker, docker compose, Packer

• Worked for Audience Studio which helps gather customer data, target the right audience, and make advertisement more efficient.

• Primarily worked & owned the Audience Studio console backend API application written in Python, Django, & Tastypie with Amazon RDS + Redshift as the persistence and Redis as cache

• Lots of TDD + Docker

Test Parallelization:

• Improved CI build time of unit + integration test run time of console from 65 mins to 17 mins (4x improvement) with multiprocessing on 4 cores

• Resolved numerous concurrency issues with tests and wrote guidelines on concurrent safe testing with Django

Dockerization:

• Successfully lead the switch from vagrant to docker for local application build/run using docker compose

• Utilized Hashicorp Packer to create docker images and push to ECR from a Jenkins build

• Switched Jenkins build for unit + integration test to utilize docker images and run tests against local MySQL + Postgres containers instead of AWS RDS + Redshift which saved $$$$

• Build time further improved by 40% (17 mins to 10 mins) on 4 cores

Library Upgrades:

• Lead a team of 5 engineers across different TZ

• Successfully upgraded from python2 to python3

• Successfully upgraded Django 1.11 to 2.2

• Ensured minimal disruption to customer by running upgraded application in canary mode

• Resolved issues related to password hashing & redis cache ser/deser to ensure regular & canary instance can run compatible to each other

Features delivered:

• Worked on Einstein Segmentation which helps find top n segments of an organization from customer

data

• Worked on offline file imports which helped ingest 1st party customer data of organizations

• Improved backend job copying 3 TB partner data from 8 to 3.5 hours

Mentorship:

Mentored interns & junior engineers on good software development practices

• Barkly is a SaaS based endpoint security solution provider.

• Learned the art of TDD and good software engineering practices thanks to some excellent mentorship & smart colleagues.

Malware Incident management:

• Designed and implemented a fault tolerant solution for grouping incidents for a better customer

experience.

• Improved the average load time of large number of incident groups by 50% over the old implementation.

Rapidvisor™ upgrade:

• Implemented solution for managing manual & auto upgrades of endpoints Rapidvisor™ version

Billing:

• Designed billing for organizations to manage subscriptions & adjustments sensitive to the number of endpoints

added/removed using Recurly.

• Added feature to download invoice pdf and fixed bugs in the open source Recurly Java library.

https://github.com/killbilling/recurly-java-library/pull/128

https://github.com/killbilling/recurly-java-library/pull/201

API Integrations:

• Designed REST APIs to manage protection, status, upgrades and incidents for organization endpoints.

• Designed REST APIs to manage teams, endpoint protection & status, and incidents for organizations.

Lab testing:

• Implemented test executioner in Python to run test suites against Barkly agent in various Windows OS machines agent before shipping to customers.

Developer Productivity:

• Improved local redeployment time of API microservice from 4 minutes to 35 seconds.