Software Engineer, AI

The Browser Company

Saas

10-50

New York, NY, USA

About the job

Overview:

Hi, we're The Browser Company 👋 and we're building a better way to use the internet.
Browsers are unique in that they are one of the only pieces of software that you share with your parents as well as your kids. Which makes sense, they're our doorway to the most important things — through them we socialize with loved ones, work on our passion projects, and explore our curiosities. But on their own, they don’t actually do a whole lot, they’re kind of just there. They don’t help us organize our messy lives or make it easier to compose our ideas. We believe that the browser could do so much more — it can empower and support the amazing things we do on the internet. That’s why we’re building one: a browser that can help us grow, create, and stay curious.
To accomplish this lofty task, we’re building a diverse team of people from different backgrounds and experiences. This isn’t optional, it’s crucial to our mission, as we need a wide range of perspectives to challenge our assumptions and shape our browser through a bold, creative lens.
With that in mind, we especially encourage women, people of color, and others from historically marginalized groups to apply.

About The Role
Browsers know everything about us and what we do everyday, and yet they can’t predict our next move, morph themselves to better suit our tasks, or proactively take work off our plate. As the first AI Engineer at The Browser Company, you will work with product engineers to prototype and explore how we can build a smarter, more personalized web browser with a focus on privacy-preserving, on-device models.
You’ll work to answers questions like –

What kind of features are better suited for on-device models vs modern LLM APIs like GPT?
How can we get on-device inference quality to match that of GPT 3.5 or 4?
How can we get local LLM and embedding models to run faster and more performantly on our members’ machines?
Can we distill larger models into a smaller footprint? Or fine-tune local models to work well for our particular use-cases?
Can we fine-tune performant neural networks to do narrow tasks where the generalizability of LLMs are not necessary?
How do we collect or build synthetic datasets for our models in a privacy-safe way so we can continue to be one of the most privacy-sensitive browsers on the market

Overall you will...

Scope and spearhead projects to fine-tune, distill, or train models for various features within Arc
Push the boundaries of what on-device and privacy-safe AI and ML can be used for
Work with Product Engineers, Product Designers, and Design Engineers to understand how we can use heuristics, neural networks, and LLMs to create magical experiences within Arc
Build infrastructure to collect or generate training data for building or improving models
Build ways for us to determine and track model performance and accuracy, and improve performance and accuracy over time

After 1 month you will...

Onboard to the team and codebase with your onboarding buddy
Attend a number of onboarding presentations on the company, product, codebase, and culture
Get familiar with the Swift language, the Arc codebase, and how we ship features
Discuss and start formulating our ML roadmap with our CTO
Ship a few bug fixes and small improvements across our codebase and tooling
Have pair programmed with a few people on the engineering team
Be regularly posting product feedback about the browser in our #dogfooding channel

After 3 months you will...

Be familiar with how we prototype and build new features, working with product engineers to brainstorm ways to use models to add intelligence to Arc
Be familiar with our infrastructure and data pipelines
Ship a few prototypes with existing, on-device models to test performance and viability
Participate in product brainstorms to think about the future of Arc
Regularly attend weekly engineering discussions about our architecture, how we do code review, code style, and more

After 6 months you will...

Creatively solve problems with product engineers, using pragmatic solutions ranging from basic heuristics, regressions, ML models, to AI depending on the feature
Drive projects from conception to production launch independently
Own our infrastructure to collect training data and fine-tune models for our use-cases
Have built out mechanisms to assess quality and performance, and be working with product teams to improve the efficacy of our models and heuristics
Be interview trained and interviewing candidates for roles at the Browser Company
Be mentoring and pair-programming with newer engineers to help them get spun up on the codebase

Qualifications

You have experience developing and productionizing modern machine learning models, especially ones that can run under tight power and performance constraints, in a real-world product environment
You have experience with fine-tuning, distilling, and improving existing ML models
You're passionate about prototyping product features and using pragmatic solutions to solve problems, biasing for fast feedback loops
You’re experienced with Python and modern ML libraries
You have experience as a technical lead on critical projects or initiatives within your team and organization
You're pragmatic and can see the bigger picture. You're able to reason about prioritization and scope
You have a bias for action. We like learning as quickly as possible, so we embrace failing fast to refine our code, systems, and processes rapidly