Alexander Yevchenko

per aspera ad astra

Projects

Long-term work, research, and things built under pressure.

Cultura

2024 · Co-founder

Cultura

“Crop phenology for farmers who can't afford to be wrong.”

Co-founded an app delivering crop phenology and financial tools to small-scale farmers. It started as a hackathon project, placed in Buildspace's S5 Top 32, and was presented at WebSummit Lisbon and SXSW. Backed by Emergent Ventures and The Iris Project.

Backing

Emergent Ventures · Iris Project · Rideau Hall Foundation

React Native · ML · Agriculture · Startup

fundme

2026 · Builder

fundme

A members-only funding directory for Engineering@Carolina: every grant, investor, and pitch competition a student founder at UNC might need, kept current by the club. Grants and investors are split by whether they take equity, pitch competitions cut across both, and only members an admin has approved can see any of it.

Next.js · TypeScript · PostgreSQL · Supabase Auth

Collimator

2023 · Co-host

Collimator

Co-host of a podcast bringing together leaders in business, technology, and science to find the signal in the noise. 30+ episodes to date with guests ranging from founders to professors to investors like Harry Gandhi of 1517 Fund.

Podcast · Media · Interviews

UNC Applied Engineering Lab

2024 · Student Researcher

UNC Applied Engineering Lab

Developing a fully-waterproof, pose and movement-tracking glove for swimmers, designed to optimize hand positioning and improve performance for UNC athletes.

Hardware · Signal Processing · Sports Tech

GPT from Scratch

2024 · Self-directed

GPT from Scratch

Rebuilt a Shakespearean LLM to understand the architecture of GPT models, following Andrej Karpathy's curriculum: a character-level, 6-layer, 6-head transformer with 384-dimensional embeddings and a 256-character context, trained from scratch in PyTorch on Shakespeare's text.

Python · PyTorch · NLP · Transformers

Corn Disease Computer Vision

2023 · Independent Research

Corn Disease Computer Vision

Researched food supply pressures from a first-principles perspective. Developed a computer vision algorithm for drones to identify and treat corn disease in real-time. A PyTorch CNN trained on PlantVillage leaf images tells healthy leaves apart from common rust, northern leaf blight, and gray leaf spot with about 92% test accuracy.

Computer Vision · Python · Agriculture

Hackathons