Atomic is a venture fund that founds companies. Founded in 2012, we believe that disruptive innovation is most successfully achieved by pairing innovative ideas with business discipline, and that building those ideas into businesses is not something that can be outsourced. We are engineers and entrepreneurs who build and operate the next generation of great companies.

As one of Atomic's pre-launch startups, we are building the payments stack for the modern subscription economy.

Let's get to the point. You're our machine learning systems builder. You use technology to make training and deploying models as seamless as possible, because human time is much more expensive than machine time.

A well functioning training and deployment environment accelerates the feedback loop and time-to-market for ML algorithms. We can't forget about observability either. 

Deploying machine learning models in production requires thinking and working with non-deterministic output where the grittiness is in the data (potentially unversioned), not the code.

This has inspired a new generation of platforms for deployment and ops but they are still a few years away from completely solving this problem space. We need someone who lives and breathes containerization, scaling, distributed systems and enjoys diving into algorithms as well, since optimizing an ML algorithms comes with nuanced tradeoffs around accuracy and performance. In this case performance can be ambiguous - it is training time, inference cost, inference time, or deployment time. You're here to help us define and balance these tradeoffs.

You'll get to architect our system and lay the foundation for the future from both a technology and a system design perspective. Once you're done we'll have no problem seamlessly deploying multiple models a day, being made aware of drift as soon as it happens and falling back to old models when new ones unexpectedly fail. You'll work closely with our our Eng team to ensure that we optimize the design of the entire system and make good tradeoffs along the way.

You'll also collaborate on the algorithm side, as it never hurt to have another collaborator, much in the same way that Random Forests outperform Decision Trees. 

You're familiar with distributed systems and data processing frameworks such as Scala and Spark. You've used ML systems such as sklearn, tensorflow and/or pytorch. Yes, other languages/frameworks are fine as long as you can articulate how you've used the technology to solve similar problems. You don't mind writing SQL as necessary, but probably wouldn't operationalize parameterized notebooks, even if you could. 

Matrix multiplication is something you dream about. I'm only half kidding here as it underpins just about every aspect of modern ml and contains the keys to optimizing any algorithm using it. 

Designing an ML deployment system from scratch is something that excites you and you're able to clearly articulate the tradeoffs of different approaches and platforms. You've likely built some version of this before, or are at least intrigued by the idea of serverless inferencing.

You strongly believe that action creates information.

You want to work on a small team and have lots of responsibility.

You look forward to being scrappy and enjoy overcoming challenges.

You are willing to admit when you don't know something and bring in others to help you figure it out.

We are focused on building a diverse and inclusive workforce. If you’re excited about this role, but do not meet 100% of the qualifications listed above, we encourage you to apply.

-----

Atomic is an Equal Opportunity Employer and considers applicants for employment without regard to race, color, religion, sex, orientation, national origin, age, disability, genetics or any other basis forbidden under federal, state, or local law. Atomic considers all qualified applicants in accordance with the San Francisco Fair Chance Ordinance.

Please review our CCPA policies here.