Software engineer at Lyft

Swapnik Shah

Backend systems people rely on, for students, AWS customers, and Lyft drivers.

I build backend services on Lyft's driver team and run experiments with product managers and data scientists to see what actually helps drivers. Before Lyft, I worked on AWS IAM Identity Center and spent eight years building student tools at the University of Minnesota.

Work

  1. Lyft

    2025 – now

    Software Engineer, Driver team

    • Server lead on ride auto-accept, which accepts rides for drivers who opt in. Backend across four services, from a small beta to full rollout.
    • Led the backend that spots drivers who accept a ride but don't head to the rider. It combines two ML models owned by other teams, forgives one-off misses, and warns before it penalizes.
    • Built the team's end-to-end test framework against staging.
  2. Amazon Web Services

    2023 – 2025

    Software Development Engineer 2, IAM Identity Center

    • Owned control plane APIs for IAM Identity Center across 25+ regions.
    • Brought the service to new regions and helped launch account-level instances.
    • Led four engineers on a GenAI code reviewer built on Bedrock. It cut revisions per review by about 20% and won an org hackathon.
  3. University of Minnesota

    2015 – 2023

    Software Developer

    • Built the advising system used by 60,000 students and thousands of staff across all five campuses.
    • Trained a model that flags students at risk of falling off track, so advisers can step in early.
    • Designed 50+ REST APIs and moved legacy apps to modern Laravel and Docker with no downtime.
  4. Mississippi State University

    2011 – 2015

    Web Developer

    • Built websites, databases, and data visualizations for campus departments.

How I work

  1. 01

    Ship it as an experiment.

    Ramp slowly, watch the numbers, and roll back when something looks off.

  2. 02

    Use AI, check everything.

    I use Claude Code daily with my own skills and hooks, so every change runs checks and staging tests. AI-written code gets the same review as a junior engineer's PR.

  3. 03

    Spec first.

    Map how data is read and written, and tell downstream teams what load to expect, before writing code.

Tools

CLI · Python

prompt-gen

Packs a codebase into one prompt-ready file, with gitignore-aware filters and size limits.

In progress

Trip planner

Plans a road trip around the constraints your group actually has.

Next, I'm working toward launching a few AI products of my own.

Games

Word game

Quintet

Guess the five-letter word in six tries. Color hints, stats, and a dark mode.

Up next

More games on the way

Small games that run right in your browser. No installs, no accounts.

About

I've been a software engineer for over 10 years, at two universities, AWS, and Lyft.

Education

  • MBACarlson School of Management, University of Minnesota
  • Post Baccalaureate Certificate in Data ScienceUniversity of Minnesota
  • BS in Computer Science, Mathematics minorMississippi State University

Outside work

  • Snowboarding
  • Making lattes
  • Sim racing