austin berg.

full-stack data & ml uva '27

austin berg.

CS & Applied Statistics student at UVA and founding engineer at GreekCore. I build full-stack products, backend systems, and ML pipelines, from raw idea to production.

see the work
résumé
githublinkedinx
  • TypeScript
  • Python
  • Go
  • React
  • NestJS
  • FastAPI
  • PostgreSQL
  • Redis
  • Stripe
  • Docker
  • Kubernetes
  • GCP
  • Terraform
  • XGBoost
  • PyTorch

by the numbers

$0m+

processed through the payments layer i helped build at greekcore

0+

members using greekcore daily, up from 1,500

0k+

mlb pitches piped through my etl + ml pipeline

0+

chapters on the platform nationwide

01

about.

Outside of code, I'm the president of the Beta Psi chapter of Sigma Alpha Mu, secretary of Google Developer Groups @ UVA, assistant coach for the UVA club hockey team, and a team leader in UVA's chapter of Blossom Together.

the toolkit

languages

PythonTypeScriptSQLRGo

frameworks & libraries

ReactNestJSNode.jsFastAPISocket.IODrizzle ORMBullMQ

ml & data

PyTorchScikit-LearnXGBoostPandasDuckDBPrefect

cloud & infrastructure

GCPCloudflarePostgreSQL (Neon)RedisDockerKubernetesTerraform

tools & practices

Claude CodeGitHub ActionsGitOpenTelemetryStripe APIs

02

experience.

~/austin/career · git log --oneline --graph
  • a13f9c2(HEAD -> greekcore/founding-engineer)Jan 2026 – Present

    Founding Engineer @ GreekCore

    • Co-led a ground-up rebuild of a production SaaS platform as one of two engineers, re-architecting core infrastructure and rebuilding real-time chat used daily by 3,000+ members.
    • Engineered a custom payments layer with Stripe financial APIs that has processed $1M+ in transaction volume nationally, implementing idempotent payment flows for financial data integrity.
    • Migrated the full production codebase and database from Replit to a GCP microservices ecosystem with Neon (Postgres) and automated CI/CD via GitHub Actions, cutting infrastructure costs by ~$10.8K annually.
    • Shipped features and maintained on-call production support as the platform scaled from 1,500 to 3,000 members across 45+ chapters and grew MRR from $3K to $7K (+133%) in five months.
  • 7e2b4d1(pelagic/swe-intern)Jun 2026 – Present

    Software Engineer Intern @ Pelagic AI

    • Contribute production code to a polyglot network observability platform (Go, TypeScript, Python, Vue) used by defense and intelligence customers including Maxar and Jacobs.
    • Deploy and operate containerized services on Kubernetes using Helm charts, Traefik ingress, and Docker Compose.
  • 3c8e5f9(pelagic/ai-ml-intern)Jun 2025 – Aug 2025

    AI/ML Intern @ Pelagic AI

    • Designed a three-table SQLite schema, reducing scenario initialization time by 50%.
    • Refactored FastAPI endpoints and centralized logging to improve latency and consistency.
    • Built an OSM airport graph generation tool, live-demoed to Pelagic, proving a concept they later implemented.
  • 9d4a7b3(pelagic/ml-intern)Jun 2024 – Aug 2024

    Machine Learning Intern @ Pelagic AI

    • Trained machine learning models on orbital-mechanics data, predicting satellite slew time and reducing error.
    • Engineered datasets and visualizations to enhance model accuracy and efficiency.
    • Delivered actionable insights and production-ready code to Jacobs' $16B Aerospace division.

tags:

supervised-mlSupervised Machine Learning: Regression and Classificationadvanced-algorithmsAdvanced Learning Algorithmsunsupervised-rlUnsupervised Learning, Recommenders, Reinforcement Learning
03

selected work.

01 / 03featured2026

GreekCore Platform.

Ground-up rebuild of a production SaaS platform serving 3,000+ members across 45+ chapters: real-time chat, a custom Stripe payments layer with $1M+ processed, and a full migration to a GCP microservices ecosystem.

TypeScriptReactNestJSStripeGCPNeonRedis

02 / 03featured2025

MLB Pitch Prediction Pipeline.

End-to-end ETL pipeline ingesting, transforming, and validating 140,000+ MLB pitches with Prefect & DuckDB. Engineered 20+ features, deployed a FastAPI service with live ETL triggering, and trained XGBoost & Random Forest classifiers achieving 72% fastball accuracy across 9 pitch classes.

PythonDuckDBPrefectFastAPIXGBoostDocker

03 / 032026

This Very Website.

The site you're looking at, designed and built from scratch with Next.js 15, Framer Motion, and Tailwind CSS. Featuring an interactive dot garden, because why not.

Next.jsTypeScriptFramer MotionTailwind CSS

off the clock

Ice Hockey

Watching, playing, coaching, you name it.

Climbing

Bouldering and top roping, of course.

Skiing

As long as there's fresh powder.

EDM & House Shows

Catching live sets whenever possible.

Sailing

On the water whenever the wind cooperates.

Guitar

Still figuring out bar chords, but I'll get there.

Gaming

Strategy games and neat open-world adventures.