Volvo 2026 Internship

Maintenance Cost Calculation Tool 2.0

Converting an old microsoft powerapp to a modern python webapp

Summary:
Tech Stack: Python (Nicegui, SqlAlchemy), Databricks, Azure services like Azure Devops (and their .yml format) Azure Entra, MSSQL Server and Azure Data Factory. I got accepted into a Janurary 2026 co-op position in the Engineering Maintainability team. My job involved some data analyst work (making powerBIs, reports) with a lot of software engineering work.
The biggest project I worked on was the Maintenance Cost Calculation Tool 2.0. I basically took this slow, buggy powerapp that was used by 40 Maintainability Engineers across 5 countries and made it into a modern Python Webapp.
On top of creating the app in a new technology and framework, I spent a lot of time on optimizing costs speed. The speed increase ranges from 5x-50x. I got this by making a lot of calls asynchronous (so when we have a page that needs 10 sql queries, we do 10 at once instead of doing them 1 by 1) and by using caching whenever I could (nicegui has user.storage.client for in-browser storage).
We also have massive cost improvements too. I spent a lot of time making our notebooks that are ran manually into jobs that are ran automatically on weak, cheap clusters overnight. I also made it so we used Volvo's shared compute rather than provisioned resources from Microsoft. Our biggest win was turning a ~30 minute job into one that was ~4 hours at 4% of the cost.

DiscussDock

Creating a full stack social media website Spring-Summer 2025.

Social media is toxic and full of tracking. People cannot express themselves and ask honest questions. I want to fix that.

Summary:
Tech stack: Docker, Github actions, Go, Html + CSS (Tailwind css, DaisyUI), Python (llm finetuning, data set collection), Mongodb (atlas)
Complex Objective: Create an anonymous social media website that utilizes a fine tuned LLM to reduce toxicity and enforce professionalism.
Status: Mostly done. Server costs are getting high, and I'm out of google cloud tokens. The site is in maintenence mode (9/12/2025) and I'm deep into grad school. Because it's anonymous, it's hard for me to get donations and to form a community.
Key takeaways: I can build a lot. I can make something to fix a real life problem. However - an overwhelming majority of social media users are mobile, and they have an important set of expectations. If I don't have an explosive set of users up front, it's hard to have users trickle in over time.
Make sure to have people you think are cool check out the site! I've got so much help developing this from family and friends. It's good to branch out to non-technical people, and see what they think aobut your work. I'm a strong believer of "diversity is a strength". Get diverse viewpoints!
Check it out here!

ERC - HAMMER Digital Twin Development

Creating a digital twin backend Fall 2023 - Fall 2024

In this project, I led a team of around 3-5 students to create an extremely complex AI-powered digital twin for English Wheel Forming.

To summarize:
Tech stack: Docker, Github actions, Go, Html + CSS, Python, Unity3d
Complex Objective: Create a bi-directional low-latency real time communication backend that is able to support multiple concurrent users, disparate file formats with secure data transfer.
Status: Successful, with a scientific elsevier paper awaiting publication (as of 12/27/2024)
Key takeaways: I now understand what it takes to create robust full-stack applications. Performance matters.

Goober's strange escape - a video game (and my senior project)

Creating high-quality real-time enemy AI Fall 2023-Spring 2024.

Tech stack: C#, Google Cloud, Unity3D.

This game is a first person co-operative wave shooter about playing a group of test subjects who are fighting against a bunch of monsters.

My responsibility: Implementing Goal-Oriented-Action-Planning into a real-time networked video game.

This was difficult. The AI worked by analyzing the state of the world, and then making plans to maximize damage dealt to the player.

Enemy patrols a zone, uses a collision detection algorithm and randomized directional movement algorithm to find an area that's open.
After a few cycles of this, the enemy will randomly decide to: Get a better gun, use their ears to hunt down the player, or group up with its allies.
Once a player is spotted, the enemy will move quickly alongside a circumference around the player.
The enemy will strafe and shoot until the player dies.
If they run out of ammo, they will either try and melee the player or run for more ammo (depending on health amounts, and a random value)
If they get hurt, they get a teleportation ability and run away to get health.

Key takeaways: How real-time, multithreaded game AI works. How to work as a democratic team.

Play it here!

Energy usage simulation - Revit & Unity3d

Modelling energy usage Summer 2022

My team of three used historical and synthetic datasets on a CAD (Revit) model to simulate energy usage/generation for a modern eco-friendly home. We then took all of this data and placed it into Unity3d, where all it was visualized in a real-time and interactive environment.

To summarize:
Tech stack: Revit, Python, Matlab, Kaggle datasets (energy usage & generation), Unity3d
Status: Successful, with plans to expand with further Key takeaways: This was my first time working with synthetic datasets, and I understand how/why they are so commonly used. This is also where I first understood the value of putting things in interactive 3d environments.

Watch it here! (Google Drive Link)