Magnus Johnson

Hello! I'm a software engineer based in San Francisco with expertise in 3D, machine learning, computer graphics, design and leadership. I graduated from MIT with bachelor's and master's degrees in Electrical Engineering and Computer Science in 2020. I'm also passionate about education, music, art and technology.

I've spent 6+ years at Apple working in the Vision Products Group, 12+ years shipping independent projects under MagMHJ, and held many jobs & internships at MIT, Cognite, Vivint, Uber and a handful of other companies.
Resume Email LinkedIn

Portfolio

I drove engineering for several Spatial Preview features (immersive preview, lighting effects, gesture controls) and coordinated efforts across design, QA and engineering teams. Many of these features shipped in visionOS 27.

I also developed management tools for our Spatial Preview presubmission pipeline to accelerate cross-functional QA/developer triage on urgent issues, increase build acceptance rate and improve product stability.

I led engineering efforts for Persona Lighting Filters on visionOS, which improved user acceptance of Persona avatars above a critical launch threshold. This feature has been showcased in Apple marketing materials and product announcements.

I also designed and productionized a light normalization ML model and inference pipeline for visionOS Persona enrollment to expand robustness across real-world lighting scenarios.
I also designed and shipped a novel don/doff recognition algorithm using signal processing techniques, which resolved a critical rendering issue blocking Personas release.

Though many of my ML research projects at Apple have not yet been published, a few have, including these:

Devices and methods for generating and/or presenting three-dimensional landscapes

This project allows users to place their own photos within a blank panorama, and the system will generate a 3D environment to bring the user's photos "to life" around them and blend together the photos with realistic environmental details.
The system consists of many independent models, each trained with separate data, objective functions, and methodologies. These models include a segmentation network, several generative infill networks, a fluvial line simulation model, and a foliage simulation model.

Point cloud enhancement using an infill mask and synthesized representation

This project addresses the incompleteness of captured point clouds, where occlusion and poorly reflecting surfaces leave holes and sparse regions that degrade the output render quality. The goal of the system is to densify an input point cloud captured from a single perspective using generative methods.

Lighting normalization for Personas enrollment

This project involved the creation of a specialized machine learning model to normalize lighting conditions during Persona enrollment. The architecture and design of this system are proprietary and cannot be discussed with the public, but the system has shipped in visionOS and has been demonstrated in marketing materials.

In 2020, I released two apps, MINIMA Sculptor and MINIMA Animator, as accessible-but-powerful creative tools for 3D modeling and animation. Since then, they have been downloaded over 54.5K times, featured in educational stores, and highlighted in several newsletters.

MINIMA Sculptor

MINIMA Sculptor lets users construct 3D models by combining and fine-tuning properties of 3D primitives. I designed it to be as simple as possible so it would make 3D modeling accessible for kids and novices.

MINIMA Animator

MINIMA Animator is a simple, powerful 3D animation tool that integrates directly with MINIMA Sculptor. It shares the same accessibility goals as MINIMA Sculptor, but "digs deeper" and offers more detailed controls for more technical projects.

Before joining Apple, I released several apps as the founder and sole member of MagMHJ, LLC. Many of those apps are now out of date or no longer available, but I continue to maintain MagMHJ.com as an archive of this experimental early period of my career.
Here are some of my favorite legacy MagMHJ projects:

mobileML

mobileML is a neural network playground I released in 2018 built upon a custom neural networking framework that I wrote from scratch. Users could design, train, and run their own local networks with live sensor data from their mobile devices.

Archipelago

Archipelago is a full-length survival video game I created in 2015 with my own custom 3D rendering & physics engine. It features several complex mob AI systems, a detailed weather system, and a simple crafting system. It was one of my most ambitious, rewarding and fun side projects to date.

TapNote

TapNote was the first educational tool I ever publicly released in 2014. It was a surprise success on the App Store and ignited my passion for making creative tools.

I've been fortunate to have some of my work at Apple be published as patents:

Granted

Pending

Interests

I love many types of design, but I particularly enjoy visual design (motion graphics, shaders, visual effects, etc). I like to experiment with real-time shading/rendering designs in my spare time, like these:
I also love to make more abstract designs, usually for marketing purposes or wallpapers:
I also periodically dabble in crafts, like paper foldables:

I care a lot about education, and I've held a number of teaching roles throughout my career, including teaching music classes at MIT, teaching swim lessons, and tutoring high school students.

In college, I founded the MIT Animation Group to teach animation practices that were not covered in class coursework. I've also led a number of volunteer educational workshops to teach music composition with TapNote, a music composition app I released in 2014.
I also try to incorporate educational aspects into my software projects and interfaces. My goal is to share my own learnings with others through teaching and accessible design principles.

In my spare time, I like to write music and make remixes of movies and video games. One of my happiest memories is getting a shoutout from POGO on one of his livestreams for my song "Hey Buzz":

I love spending time outdoors hiking and backpacking with my family. Here are pictures from some of my favorite hikes from the past few years: