Hi, my name is

Altti Koskinen.

Applied & Computational Math Sciences student at the University of Washington, currently building real-time GPS tracking at Webscorer Inc., with skills in full-stack development, data science, and machine learning.

01. About

I'm pursuing a B.S. in Applied and Computational Math Sciences at the University of Washington, with minors in Applied Math, Data Science, and Informatics. I love building projects that turn math and data into things people can use.

A few technologies I've been working with:

  • Python
  • C++
  • Java
  • TypeScript
  • SQL
  • PyTorch
  • FastAPI
  • React
  • Node.js

02. Experience

Software Engineer Intern @ Webscorer Inc.

Mar 2026 – Present · Mercer Island, WA

  • Developing a real-time Android GPS map and tracking module with the Google Maps API, delivering 1Hz location updates.
  • Recording and displaying racer route, and supporting full race replay functionality.

03. Things I've built

Cloud Radiative Effect Deep Learning Research

Jan 2026 – Mar 2026

Worked with ESS graduate students to build a CNN climate model in PyTorch predicting longwave and shortwave Cloud Radiative Effect from global sea surface temperature patterns using CMIP CanESM ensemble data. Engineered an end-to-end pipeline for NetCDF preprocessing, dataset concatenation spanning 250 years, and CRE derivation, training a model with average skill score of 91.14% across 275M+ data points.

  • PyTorch
  • NetCDF
  • CNN
  • Climate Science

Conversational Chrome Extension

Dec 2025 – Feb 2026

Built a Tutor and Mock Interview Chrome extension with JavaScript, scanning current tab data for context, with a FastAPI backend delivering real-time conversation using Kokoro for TTS audio generation. Used Llama 3.2 to generate responses, hosting a server handling real-time voice processing and response generation — a fully offline conversational agent achieving ~15s response times.

  • JavaScript
  • FastAPI
  • Llama
  • TTS
GIF

Spectral Simulation of 2D Vortex Dynamics

AMATH · Scientific Computing

Solved the 2D vorticity–streamfunction form of the Navier–Stokes equations on a periodic grid using an FFT spectral method and solve_ivp time integration. Simulated colliding vortex pairs, like- and oppositely-signed vortices, and a turbulent field of many interacting vortices — rendered as animations.

  • NumPy
  • SciPy
  • FFT / Spectral
  • Fluid Dynamics
View details →
PDF

The Dynamics of Love through Mathematics

Written Paper · Differential Equations

A paper modeling romantic relationships as coupled systems of differential equations. Explores how feedback between partners' feelings produces stable, oscillating, or runaway dynamics, with phase-plane and stability analysis of the resulting fixed points.

  • ODEs
  • Dynamical Systems
  • Stability Analysis
Read paper →
PDF

PyTorch CNN Manual

Technical Manual · Deep Learning

A from-scratch instructional manual walking through building and training convolutional neural networks in PyTorch — covering tensors, convolution and pooling layers, the training loop, loss and optimization, and practical tips for getting a model to converge.

  • PyTorch
  • CNNs
  • Deep Learning
Read manual →

Go To Class!

DubHacks 2025

A mobile app that gamifies class attendance with a weekly leaderboard among friend groups. Built with Expo React Native and a Firebase backend for auth and data, turning showing up to class into a friendly competition with streaks and rewards.

  • React Native
  • Expo
  • TypeScript
  • Firebase

Exploratory Data Analyses

Data Science

A growing collection of end-to-end EDA notebooks spanning topics like AI & productivity and education & career success. Each walks through cleaning and preprocessing, univariate to multivariate analysis, and visualization-driven conclusions.

  • pandas
  • seaborn
  • scikit-learn
  • Jupyter

NBA Clutch Win Predictor

R · Shiny App

An interactive Shiny app analyzing 39,900+ clutch shots since 1996–97 to study who performs under pressure and how it shifts win probability. Visualizes shot efficiency, per-game win-prob impact, and how attempts and experience relate to clutch success.

  • R
  • Shiny
  • Statistics
  • Data Viz

What's next?

Get in touch

I'm always open to new opportunities and I love building and solving problems. Whether you have a role in mind or just want to chat, my inbox is open.

Say hello