KARSTEN
LOWE

I build intelligent software end to end, from ML models and LLM-powered tools to the full-stack apps that serve them.

View work

About

A creator and innovator of data and software solutions, studying Data Science at UC San Diego.

I'm a Data Science student at UC San Diego who takes ideas and turns them into reality for people to actually use.

Across roles at GelTech Labs, Laxcen Technologies, and UCSD Athletics, I've shipped machine-learning models, real-time cloud pipelines, and full-stack apps that take raw sensor or sports data all the way to a clear decision.

My through-line is end-to-end ownership: I build the data layer (PostgreSQL, MQTT, AWS), the models (XGBoost, PyTorch, scikit-learn), and the product around them (React, Next.js, FastAPI), increasingly wiring LLM and agentic workflows on top. I'm an AWS Certified Solutions Architect, and I care most about work that's meaningful. I'm always excited to take on a challenging problem to solve.

Data Science Student · UC San Diego

Python · PyTorch · React/Next.js · AWS · LLMs

  • 3rd UCSD DataHacks 2026
  • AWS Certified Solutions Architect
  • 15+ Tutoring clients served

Work Experience

  1. June 2026 – Present

    Machine Learning & AI Engineer

    GelTech Labs · San Diego, CA

    • Developing AnGel, a cloud-hosted prediction service that collapses GelTech's multi-week biomaterial assay cycle into a seconds-long query, letting technicians cut low-viability formulations before spending bench time and reagents.
    • Researched a multi-model reconciliation architecture (conformal intervals, kNN OOD detection, missing-feature and importance signals) that fuses predictions via inverse-variance weighting and a local Llama 3.2 reasoning layer, benchmarked on a ground-truth harness to establish where statistical vs. LLM fusion wins.
    • Engineered an LLM-powered ingestion pipeline (Python + Next.js) that parses open-source research repos into a standardized ML training corpus with human-in-the-loop expert review, cutting manual curation per study from 4 hours to 5 minutes.
  2. June 2026 – Present

    Software Engineer

    Laxcen Technologies · Laguna Hills, CA

    • Built a shipment tracking tool giving logistics live visibility into 2M+ RFID inlays in international transit, replacing email threads that had no aggregate view of in-flight inventory; containerized with Docker on AWS ap-east-1 for reliable overseas access, with Auth.js authentication and automated status reminders via Microsoft Graph.
    • Shipped an internal pricing tool for the sales department, cutting 1–2 days from each of 100+ quotes by pairing a PostgreSQL price database with the Frankfurter API for live USD/CNY rates to quote North American customers in accurate USD.
    • Overhauled technical SEO across Laxcen's 76-page site: meta descriptions, HTML parsability, and crawl errors.
  3. January 2026 – Present

    Data Analyst

    UCSD Women's Volleyball (NCAA Division I) · San Diego, CA

    • Built an in-game play predictor using recurrent neural networks (RNNs) to forecast opponents' next attack from live data.
    • Developed a Regularized Adjusted Plus-Minus (RAPM) model with scikit-learn RidgeCV over Big West Conference VolleyMetrics data (1,000+ Division I players), ranking each player's contribution to win probability.

Projects

CarbinWatcher waste-classification dashboard
April 2026 – Present

CarbinWatcher

3rd place at UCSD DataHacks 2026: an autoscaling cloud backend for a computer-vision smart trash can that tells users which bin to use. Streams MQTT through AWS IoT Core, Kinesis Firehose, and S3 (with parallel DynamoDB writes) into a live React/Next.js app, scaling to 1,000+ cans, with Databricks + XGBoost analytics predicting environmental impact.

  • Python
  • C++
  • React
  • AWS
Live demo → Devpost →
KL Math Prep student dashboard
March 2026 – Present

KLMathPrep

A full-stack AWS web app for my math-tutoring business: notes, Cognito authentication, student data, scheduling, and automated reminders. Python Lambda functions (triggered by EventBridge) send SES confirmation, cancellation, and reminder emails before each appointment.

  • React
  • Next.js
  • TypeScript
  • AWS
Visit site →
UCSD WVB Smart Analyst chat interface
April 2026 – June 2026

UCSD Smart Analyst

A PostgreSQL-integrated AI agent for the volleyball coaching staff: an OpenAI agent collects VolleyMetrics data through the UCSD portal, then cleans and loads it into Neon. Custom Python tools run validated SQL with schema introspection and a query-correctness check, returning team-specific stats and Plotly visualizations general-purpose LLMs can't match.

  • Python
  • React
  • PostgreSQL

Skills

Languages

  • Python
  • TypeScript
  • JavaScript
  • Java
  • C/C++
  • R
  • SQL
  • HTML/CSS

ML & AI

  • PyTorch
  • XGBoost
  • scikit-learn
  • LLMs & agentic workflows
  • RAG & function-calling
  • OpenAI API

Cloud & Data

  • AWS · Solutions Architect
  • S3
  • Lambda
  • DynamoDB
  • Kinesis
  • SES
  • Cognito
  • Amplify
  • IoT Core
  • PostgreSQL · Neon
  • Databricks
  • MQTT

Frameworks & Libraries

  • React
  • Next.js
  • Node.js
  • FastAPI
  • JUnit
  • pandas
  • NumPy
  • Matplotlib
  • seaborn
  • Plotly

Developer Tools

  • Git & GitHub
  • Docker
  • Linux
  • Vercel
  • Tableau
  • VS Code
  • PyCharm
  • IntelliJ

Education

Sep 2025 – June 2028

B.S. Data Science

University of California, San Diego

La Jolla, CA · Bachelor of Science

January 25, 2026

AWS Solutions Architect – Associate

Amazon Web Services

Cloud architecture certification

Get in touch

Let's Connect

I'm open to internships, research, and collaborations, especially in machine learning, data, and full-stack work. Send a note and I'll get back to you.

Socials