Hi, I’m Varun.
I make complex
work feel simple.

Software Development Engineer @ Amazon

VIEW MY RÉSUMÉ

PORTFOLIO / SEATTLE, WA

VARUNGANGAVARAPU.

SOFTWARE DEVELOPMENT ENGINEER

I build systems that turn complex ideas into useful tools, from optimization and data platforms to the interfaces people rely on.

OPTIMIZATION DIGITAL TWINS VECTOR SEARCH PRODUCT ENGINEERING JAVA PYTHON TYPESCRIPT JAVASCRIPT GO C++ SQL HTML CSS REACT NEXT.JS NODE.JS SPRING BOOT FLASK EXPRESS TAILWIND CSS PANDAS NUMPY PYTORCH AGENT TOOLING & RULE SETS MCP RAG VECTOR DATABASES PINECONE EMBEDDINGS AWS LAMBDA API GATEWAY DYNAMODB OPENSEARCH DOCKER KUBERNETES GIT LINUX

02 / TIMELINE

A RUNNING RECORD / 2022—NOW

THE PATH
SO FAR.

The places I’ve learned, built, and put ideas to work.

THREE SELECTED CASE STUDIES FOLLOW

01 / CURRENT WORKSEATTLE, WA · 2026—NOW

EMPLOYER / AMAZON

amazon

Software Development Engineer

I work on the systems that turn complex space allocation decisions into plans teams can use.

MIXED INTEGER PROGRAMMINGAWS BATCHAGENT TOOLING
12% better clustering5× faster runtime
Dramatically lit classical sculpture against a deep blue background
01 / THE MODEL

Make placement better.

I replaced a geometric bisection heuristic with a mixed integer program using McCormick linearization to reward adjacent placements without adding relocation cost.

12%clustering gain
02 / THE SCALE

Solve in parallel.

I moved the optimizer from one process to an AWS Batch pipeline that fans out 23 solvers sized at runtime.

5×faster runtime
03 / THE CRAFT

Turn reviews into rules.

I mined more than 15,000 review comments into verification rules for an engineering agent, making recurring failure patterns easier to catch.

500+defects found in backtesting
02 / INTERNSHIPBELLEVUE, WA · SUMMER 2025

EMPLOYER / AMAZON

amazon

Software Development Engineering Intern

I built a digital twin platform that gave operations teams a shared view of their fulfillment centers.

REACTOPENSEARCHDYNAMODBAWS LAMBDA
64 centers connected5M+ rows each day
Ethereal portrait with luminous crystalline forms and motion blur
01 / THE PLATFORM

One view of operations.

I built a serverless digital twin platform for real-time operational visibility across Amazon fulfillment centers.

64centers connected
02 / THE INTERFACE

Make the data usable.

An authenticated Cloudscape React interface brought deep links, multiple views, and bidirectional URL state to embedded dashboards.

100+dashboards supported
03 / THE PIPELINE

Onboard without code changes.

Configuration-driven ingestion through S3, SQS, OSIS, OpenSearch, and DynamoDB handled 25 metrics and kept queries responsive.

5M+rows each day
03 / INDEPENDENT PROJECTHACKRU · 2024VIEW THE PROJECT

PROJECT / MUSIC DISCOVERY

LYRICAL VECTOR
SIMILARITY

Recommendation engine for music by mood

A different way to discover music: start with how a song feels, not just its genre or artist.

PYTHONPINECONEEMBEDDINGSJAVASCRIPT
Futuristic portrait with a glowing amber visor
01 / THE WIN

Won at HackRU.

Our team won the Maverick Track. I built the interactive frontend and helped take the full stack prototype from idea to a working demo.

1stMaverick Track
02 / THE IDEA

Search for a feeling.

My team built a recommendation experience that lets listeners find songs by emotional tone.

500hackathon entrants
03 / THE ENGINE

Turn music into vectors.

Spotify and Musixmatch data fed a transformer that produced 24-dimensional sentiment vectors, then Pinecone handled nearest-neighbor retrieval.

24Dsentiment space

03 / GET IN TOUCH

THE NEXT CONVERSATION

EXTEND YOUR
HAND

Tell me what you’re building and what needs to work better. I’m always up for a thoughtful conversation.