Steven Norell · Portfolio

Projects

08 Projects

Senior Capstone
Project

Ski Edge Metallurgy

Black Diamond Equipment

For my senior capstone with Black Diamond Equipment, we looked at improving the wear and corrosion life of the steel edge on their Impulse Ti 104 ski without changing the 1070 high-carbon alloy, its heat treatment, or the edge geometry. We evaluated several surface treatments and selected electroless nickel–phosphorus, which deposits a uniform coating without applied current or vacuum. We plated sample edges and characterized them against bare 1070 steel: Vickers microhardness for hardness, ASTM B117 salt-spray for corrosion, and mandrel-bend testing for coating adhesion. The coated edges showed higher surface hardness and better corrosion resistance than the bare steel.

Substrate: 1070 high-carbon steel
Coating: Electroless Ni–P

Fig. 01 · Coated edge sample ●

Fig. 02 · 3D crystal viewer ●

↗ Visit live site learnmatsci.com

Full-Stack / WebGL
Education

learnmatsci

learnmatsci.com · Live

A full-stack educational web app for materials science students. Seven interactive tools: a 3D crystal-structure viewer, a library of common crystal structures, a d-spacing and Bragg-angle calculator, an X-ray diffraction pattern generator, a Hall-Petch strength calculator, an SAED simulator, and a fusion-reaction visualizer.

Stack: React · Three.js
Infra: AWS S3 · CloudFront · Route 53 · ACM
Tools: 7 interactive simulations

Python / Geospatial
Reverse Engineering

LiDAR to Assetto Corsa

In progress · pure-Python pipeline · ~8 mi drivable

Fig. 03 · Aerial LiDAR point cloud ●

An in-progress, pure-Python pipeline that turns public airborne LiDAR of a North Carolina highway into a drivable Assetto Corsa track, no 3D editor involved anywhere. The public survey ships as tiles of roughly 15 to 30 million classified points each, and about 60 chained numpy/scipy scripts carry every tile to a finished scene, written out in the game's undocumented .kn5 binary format, which the pipeline generates by hand from a reverse-engineered spec.

The survey data has no concept of a road, so the route is derived from scratch: pavement returns are rasterized, thinned to a skeleton, and solved as a shortest path across a graph, checked against OpenStreetMap. Lane lines aren't drawn either; retroreflective paint returns high laser intensity, so the centerline tracks the actual painted stripe, with width, camber, and elevation sampled every 2 m.

Each tile runs the same fixed stage chain, parallelized across idle CPU cores rather than processed one at a time, so a 20-tile section rebuilds in about 18 minutes instead of roughly an hour run sequentially. Adjacent tiles weld together with a measured 0.00 cm seam. About 8 miles of corridor is drivable so far, with more tiles queued. Photos compare the raw LiDAR point cloud, the finished in-game track, and Street View.

Input: Classified aerial LiDAR · 15-30M points/tile
Pipeline: ~60 Python scripts · numpy/scipy · all local
Output: Hand-written .kn5 · Assetto Corsa
Status: In progress · ~8 mi continuous corridor

Aerospace / Composites
Avionics

High-Altitude Rocket

Carbon-fiber airframe · GPS payload

Designed and manufactured a high-altitude rocket reaching 3,000+ ft with integrated GPS tracking, from carbon-fiber airframe fabrication through payload-bay design and avionics integration.

Apogee: 3,000+ ft
Airframe: Carbon fiber
Systems: GPS · avionics · payload bay

Fig. 04 · Launch ●

Fig. 05 · X-tail airframe ●

Aerospace / CAD
Fabrication

RC Ducted-Fan Jet

64 mm EDF · SolidWorks · cardboard-epoxy fuselage

Designed and built a 64 mm electric ducted-fan jet from scratch in SolidWorks. The fuselage is a cardboard tube reinforced with epoxy, a low-cost composite that keeps the airframe cheap and stiff, paired with 3D-printed PLA parts and carbon-fiber wing spars. The printed components are toleranced to slip-fit the tube and mate cleanly to each other, so the airframe assembles square and bonds solidly without shimming or heavy glue fillets. A mixed four-surface X-tail drives all three flight axes without ailerons, and a GoPro payload doubles as nose ballast to hit a 25% MAC balance target.

Propulsion: 64 mm EDF
Airframe: Cardboard-epoxy tube · PLA · CF spars
Control: Four-surface X-tail · no ailerons
Balance: 25% MAC · GoPro ballast

Automotive / Powertrain
Hands-On Build

VW GTI Stage 2 Build

Clutch · downpipe · Stage 2 tune

Stage 2 powertrain build on a MK7.5 VW GTI (EA888 2.0T). Dropped the transaxle to replace the worn clutch and pressure plate, supporting the engine on a forklift through the bay while the transmission was out. On the exhaust side, fitted a high-flow catted downpipe to cut backpressure ahead of the turbo, ran an MBRP cat-back, and welded in a resonator to control drone without losing flow. Flashed a Stage 2 ECU calibration matched to the downpipe, raising boost and adjusting fueling and timing for the increased airflow. The factory limited-slip front differential (LSD) helps put the added power down to the front wheels.

Scope: Clutch · transmission out
Exhaust: Catted downpipe · MBRP cat-back · welded resonator
Software: Stage 2 ECU tune
Skills: Hands-on mechanical · tuning

Fig. 07 · In-game chat query ●

AI / Agents
Node.js

Claude_AI Minecraft Assistant

mineflayer · Claude · in-game assistant

I built a Minecraft bot that runs Claude as a player in the game, so instead of alt-tabbing to a wiki you can just ask in chat. It answers the usual game questions, like crafting recipes, mob behavior, or where to dig for an ore, and it handles things that have nothing to do with Minecraft just as easily.

The same chat also drives the bot. Ask it to follow you, guard a spot, mine, fight off mobs, change the time, hand over items, or drop in a prebuilt structure, and Claude replies with a command the bot then carries out in the world.

It's built on Node.js and mineflayer, which handle the bot's movement, pathfinding, and block and inventory control. Common instructions are matched and run locally without calling the API, so they respond instantly, and only open-ended messages get passed to the model. The model backend is swappable: a metered Anthropic API key, or a local bridge that reuses an existing Claude session to skip per-call costs.

Runtime: Node.js · mineflayer
Model: Claude · API or local bridge
Ask: Game help + general knowledge
Actions: Follow · guard · mine · combat · build
Design: Local-first, model only when needed

Composites / Aero
Fabrication

Carbon Fiber Rear Diffuser

In progress · vacuum-bagged layup

An in-progress personal fabrication project combining 3D printing with composite reinforcement: a full rear diffuser, not just a single strake. The plan is to print the diffuser body, including the floor, strakes, and end fences, on my Prusa MK3S+, then reinforce the printed core with a vacuum-bagged 2x2 twill carbon fiber skin so the composite carries the structural loads rather than the printed plastic. The main body is printed and the strakes are still being worked through.

Most of the work ahead is in the composite process: surface prep on the printed substrate for adhesion, resin wet-out and ply orientation across the diffuser's curved channels and transitions, then vacuum bagging to consolidate the laminate and pull excess resin for a stronger, low-void skin. The build currently sits between strake fabrication and the first layups, with final fitment still to come.

Core: 3D-printed · Prusa MK3S+
Skin: 2x2 twill carbon fiber
Process: Wet layup · vacuum bagging
Status: In progress · pre-fitment

Fig. 08 · Strake prototype ●