Augmented Reality Engineer
May 2022 - present
The way I typically explain my job to people is "I make Snapchat filters." The workflow is a
mixture of game and web design. I make concept art and mock behaviors for product managers, design UI, create
lens assets, and script the behavior of the lens (Snap's official name for Filters). Some lenses include user input such as touch, speech to text,
or even cues like head movements. Others use user data, remote APIs, shaders, physics and collisions,
audio or particle systems.
I started in the team responsible for making developer tools for Lens Studio: Templates, Asset Library scripts and Prefabs.
In this role, my UI had to be simple and subdued to encourage developers to mentally insert their own designs.
My code too, had to be modular and easy to understand, while providing non-coding creators with all the tools they would need
to create simpler experiences.
With time, I was also able to branch into my own creative explorations. I moved to a user experience
focused team, where my job was to work with backend and client-side developers to roll out new features. I created
prototypes as well as production-ready lenses and provided feedback on the feature. This usually included Developer
Experience Feedback and ideas for how the feature could be expanded to encompass more use cases. After the first version,
I would typically create some helper tools and example projects, so that our content teams could more easily create their
own lenses.
In terms of tooling, I used Figma to make my concept art for feature proposals Photoshop to create the art assets in my Lenses.
I would then create lenses using Lens Studio and Javascript, occasionally using Blender for some more complex 3D Graphics.
Because of the exploratory nature of my job, many of my projects are not open to be shared.
Thus, below you'll find only the projects I worked on that have since been released to the public.
My first major project at Snapchat was to create a Template for developers to use when making Snapchat Lenses - aka. Filters - using the Dialogue Machine Learning Model. This model can answer the user's question using a few paragraphs of context data. I wrote made a comprehensive example project with helper scripts and no-code customization options. In the process, I gave feedback and suggestions for future versions to the team that had made the model. I then documented both the ML model and the template itself and published the template for developers.
I handled Developer Experience for In-Lens Digital Goods: the then-upcoming feature for developers to include microtransactions in their Lenses. I scripted example projects and helper tools for early access developers to use, collected feedback, and helped the initial external developers to debug their code. Together with our client-side developers, I planned out the API calls necessary for developers to add purchasing and spending to their lenses and ensured that this important new feature was not only functional, but usable for even developers with little coding experience.
As a developer, I not only enjoy creating more complex experiences, but I think they provide more value to the user. However, the accepted wisdom in social media companies is that any increase in complexity will cause users' attention to fall off. As a result, I decided to set up an AB Test designed to test this hypothesis. Using the same art assets - Monster Costume parts, such as horns or eyes - and similar UI, I created four Lenses with slightly differing behaviors.
- The Control: no interaction. The lens turns you into a little monster.
- The Carousel: You can scroll through a couple different Monster configurations.
- Fully Configurable: You can open a menu, where you pick which assets - horns, makeup, etc. - to equip
- Fully Configurable with Saving: the same as above but you can save your configurations for later.
