Kevin Grafstrom | Software Developer

An Artistic Approach to Software Development

My name is Kevin Grafstrom. I combine my experience studying computer engineering at the University of Toronto and working as a software developer with the creativity I developed in a rock band.

👇🏻Work Experience👇🏻

Work Exp.

Work

Exp.

Full Stack Developer

(4 Month Contract)

October 2025 - Present

CarGenie AI is a car recommendation app. We make money when a user clicks a link to the car we recommend. When I joined, we had a partnership with AutoTempest, which paid us very little. I implemented the data pipeline necessary for us to work with Lotlinx instead. By doing this, marginal revenue per ad conversion increased from $0.12 to $1.14. This was crucial because the cost per conversion of our ads is around $0.70, meaning that this move made us marginally profitable. I created a dashboard to show total click conversions and clicks per user for any time period. There was another problem. The recommendations took forever to load because they used an OpenAi call. To fix this, I used OpenAI to generate a car information database so that no AI request is needed for the actual recommendation. This reduced car recommendation wait time from 8 seconds to less than one second. To save money, I implemented an email server (DKIM, DMARC, SPF) to avoid the cost of a transactional email service.

ChatKitty

Full Stack Developer

(8 Month Contract)

March 2025 - October 2025

I ported the ChatKitty application to allow embedding it in iOS, Android, and web frameworks. I also implemented a user interface in Vue and a backend in Spring for modifying the chat room membership settings. I added endpoints for asynchronous message delivery using the STOMP protocol while ensuring the accuracy of concurrent database transactions using locks with the Java Persistence API. I created an automated test suite using K6 to verify the functionality of ChatKitty's real time messaging infrastructure.

Self Employed

AI Data Trainer

May 2024 - March 2025

I compared and evaluated code outputted by AI models on www.dataannotation.tech. To verify the accuracy of the AI generated code, I executed it locally. I've verified programs written with Python, Angular + TypeScript, and PHP. Then, I wrote explanations and gave ratings for the relative quality of different AI models' coding outputs.

FORMA.AI

Software Engineer

May 2022 - Aug. 2023

I replicated a client's data pipeline from Danube into an Oracle implementation with PostgreSQL. I also reduced load times for the calculations table by 80% with React, TypeScript, and Django. When one of our clients transitioned from a data warehouse to a data lake, I adapted our internal ETL pipeline so that our results remained accurate and consistent. During this process, I built and debugged Docker containers running on Amazon Web Services EC2 and ECS. I also implemented a user interface for adding manual adjustments to commissions paid. For other clients, I wrote Spark data transformations to run on the Databricks Lakehouse Platform.

Junior QA Developer

July 2017 - Aug. 2018

I wrote more than 30 automated tests for the Indigo website using Java and Selenium. I collaborated with other developers to choose the best test cases. I also performed A/B testing on internal search ordering, increasing the search click through rate 15%. During this time, I wrote more than 20 articles on Confluence explaining our testing methodology.

👇🏻Recent Projects👇🏻

Projects

Job Application Copilot

I wrote a Python program using Selenium to automatically fill out job application forms. I still choose the jobs I apply to very carefully, verify everything, and enter the creative parts manually.

Personal Finance Analyzer

I wrote a program to parse transaction csv files from CIBC using Pandas. I exported the Pandas data frames into a PostgreSQL database. Then I created views to group the transactions into categories and calculate a total for each category.

This Portfolio Website

Other than the stock images, which I got from stockphotos.com and have a license to use, I created this website entirely by myself using Next.js, React, Tailwind CSS, TypeScript, and NGINX.

Music Trends Prediction Database

A friend and I queried the Spotify API to transfer data for the most popular artists into a PostgreSQL database. We used linear regression to determine the fastest growing artists and genres.

FPGA Processor in Verilog

In a computer hardware course I took at U of T, there was a competition that involved implementing a processor on an FPGA using Verilog. Designs were ranked by the number of assembly instructions they completed per second. My teammate and I scored second place, which gave us a bonus 4.5% on each of our final grades.

👇🏻Education👇🏻

Education

Bachelor's in Computer Engineering

Graduated June 2022

My favourite course was "Communication and Design" because we made a Google Maps clone and wrote documents explaining our design choices. This allowed us to develop our technical and writing skills in the same course. I also found " Computer Networks" very useful for understanding server and client communication in web development.