$ ./varad --init
[ OK ] mounting /portfolio
[ OK ] loading neural weights… 124M params
[ OK ] warming up CUDA cores
[ OK ] handshake with /dev/coffee
welcome.
Available for SWE / MLE roles · San Diego

Varad Kottawar.

>

Software Engineer and Machine Learning Researcher at University of California San Diego. Building intelligent systems at the intersection of artificial intelligence, distributed backends, and high-performance computing.

Varad Kottawar with a mountain dog in the Himalayas 📍 somewhere with a good dog
Scroll

What I Do

Where I add
real value.

Two years at Deutsche Bank shipping backend systems that served over 100,000 employees. Now at University of California San Diego, going deep on machine learning systems, GPU computing, and artificial intelligence pipelines. I like hard problems.

2+

Years at Deutsche Bank

156×

GPU speedup

100K+

Users served

01

Machine Learning Engineer

Transformers from scratch, retrieval-augmented generation pipelines, and CUDA-accelerated compute. Artificial intelligence built to run in the real world.

02

Backend Developer

Spring Boot microservices, REST APIs, enterprise-scale systems. 99% uptime is the baseline, not the goal. Built for 100,000+ users at Deutsche Bank.

03

GPU Programmer

CUDA kernels, parallel dynamic programming, memory coalescing. Making algorithms fly on silicon — 156 times faster than the HMMER3 baseline.

varad@portfolio: ~

Hackathons

1 entered. 1 won. 100% win rate.

Yes, the sample size is one. No, I will not be entering more until I absolutely have to defend this stat. Technically undefeated. Technically untouchable.

🏆 Winner  ·  DataHacks 2026  ·  Best Use of Marimo / Sphinx

Seismic Risk Atlas — Earthquake Loss Estimator

Built an end-to-end machine learning system estimating expected household earthquake loss across 2,498 census tracts in Los Angeles County. Combines physics-based seismic simulation (PGA, PGV, Arias intensity), HAZUS damage modeling, Monte Carlo aggregation across 500 scenarios, and an interactive choropleth risk map. Because predicting where the earth shakes and how much it costs was apparently not dramatic enough without an explainable natural language API on top.

Python XGBoost FastAPI Leaflet.js GeoPandas NVIDIA Databricks OpenAI
W

Portfolio

Selected
projects.

All repositories
01

Artificial Intelligence · Retrieval-Augmented Generation

CodeLens

A GPT-2 scale transformer built from scratch (124 million parameters) with multi-head causal attention and pre-LayerNorm residual blocks. Layered with a retrieval-augmented generation pipeline over GitHub pull request diffs for structured code reviews.

PyTorchHuggingFace ChromaDBRetrieval-Augmented Generation
02

CUDA · High Performance Computing

GPU Viterbi Aligner

↑ 156× speedup over HMMER3 baseline

GPU-accelerated genomic sequence alignment on 32 SARS-CoV-2 sequences using parallel dynamic programming and on-device backtracking to eliminate host-device memory transfers.

CUDAC++ GPU ComputingBioinformatics
03

Computer Vision · Full Stack Development

Grocery Receipt Scanner

Claude vision model extracts structured data from receipt images — line items, prices, and categories — with per-item confidence scoring, inline correction, and a spending analytics dashboard backed by a versioned REST API.

Claude VisionFlask Chart.jsSQLite
++

Always Building

More coming
very soon.

Check GitHub for the latest repositories and experiments. Hackathons, side projects, research — it is all there.

Experience

Where I have
worked.

July 2023
— August 2025

Deutsche Bank

Software Engineer

  • Architected Spring Boot microservices powering workforce analytics for over 100,000 employees at 99% uptime, with a React frontend serving 200+ senior leaders.
  • Built a low-latency Completion Suggester API for real-time search across 100,000+ records and boosted batch throughput by 40% via in-memory caching.
  • Delivered a real-time organization simulation engine enabling C-suite executives to model restructuring scenarios interactively.
May 2022
— August 2022

IoTIoT

Software Engineering Intern

  • Built a Python and OpenCV object detection pipeline to identify tumor regions in medical scans using edge detection, contour extraction, and morphological operations.
  • Tuned Gaussian blur, color space conversions, and contour filtering to improve consistency across scans with varying contrast and noise levels.

Education

Academic
background.

University of California, San Diego

Master of Science in Computer Engineering

Computer Architecture · Deep Learning Systems

3.7 2025 — 2026

Pune Institute of Computer Technology

Bachelor of Engineering in Electronics and Telecommunications

Data Structures and Algorithms · Database Systems · Object-Oriented Programming

3.8 2019 — 2023

Capabilities

Skills & Tools

Pythonexpert
Java / Spring Bootadvanced
PyTorchadvanced
CUDA / C++proficient
React / TypeScriptadvanced
AWS / Dockerproficient

Languages

PythonC++ JavaJavaScriptSQL

Frameworks

Spring BootReact Node.jsFlask

Artificial Intelligence

PyTorchTensorFlow HuggingFaceCUDAKeras

Infrastructure

AWSDocker OpenShiftMySQLCI/CD

Off the Clock

The human behind the terminal.

Current Mood (as an Engineer)

Me explaining my code
? ? ? // it works, don't ask why 🙏

Real events. The code is fine. Probably.

Contact

Got a project?
Let's talk.

Open to full-time roles, research collaborations, and interesting side projects. I respond fast.

[email protected]

Want to work together?
Let me know here.