AI Research Engineer

About Us

Poly is building a better file storage platform for everyone. We've raised $8M from YC, Bloomberg Beta, Felicis, Figma Ventures, AI Grant, and more, and we're looking to grow our founding team of superstar builders and dreamers.

Poly’s intelligent file system is a desktop and web application that lets you store and browse your files with the help of a powerful multimodal AI. An easy way to wrap your head around it is to check out a demo from Abhay

We are adamant that Poly is the future way to store files for everyone because:

1. Our product is intelligent -- all your files are indexed and searchable with a home-grown LLM that can find and even sort and organize your files, search by image and many other file types, and soon ask questions about your files and watch as an AI system references your own files to give an answer.
2. Our product is lightweight -- we don't install a sync service that's constantly scanning your hard drive and we never need to resolve sync conflicts, ever.
3. Completely, 100% cross-platform -- Our product looks identical across web and desktop, with zero compromises in the UI/UX, and all your files always available and up-to-date.
4. Our product improves every 2 days, not every 2 years -- Finder and File Explorer haven't changed in years or decades because they're linked to OS upgrade cycles. Our product ships updates every day and can add all kinds of features, fixes, and file type support at light speed.
5. Many, many more reasons -- deep folder searching, beautiful view modes, extensible metadata, ergonomics, public sharing, and so much more.

About You

We are looking to grow from two founding teammates (we've accomplished a lot as just one engineer and one designer!) to a clan of around eight, so you'll be anywhere from engineer #2 to #7. You'll be tasked with bringing Poly to life as breakthrough AI service with leading performance on relevant benchmarks, generating novel insights along the way to make Poly the “search engine for your life”.

Specifically, in a AI research engineer role, we are looking for the following concrete skills and prior experience

• High baseline knowledge of core statistical modeling principles, including the ability to write down or describe core ML modeling equations such as gradient descent, scaled dot product attention, cross-entropy, and so forth.
• Extensive experience with modeling / inference tools such as pytorch and CUDA.
• Pragmatic and product focused mindset, with the passion to see any developments make it into customer hands as quickly as possible.
• Familiarity with modern LLM-based development fundamentals including prompting, function calling, and agentic workflows.
• Knowledge of Python is a must, and enthusiasm for Rust is highly encouraged.
• An extreme appetite for learning that you've honed through recursive self-improvement. You should be able to demonstrate an exceptional aptitude for automating your own development processes.
• You thrive in independent, inter-disciplinary environments, and you are skilled at collaborating across domains and for productively engaging with people of all kinds.
• You are willing to leave judgments about gender, race, creed, background, belief, or politics for outside of the workplace so that we can focus on building the world's best file storage platform for everyone.
• You are capable of (and excited to) work in-person 5-days a week in SoMa, San Francisco

Send a friend request on LinkedIn with a short blurb (in the connection request) to me, and I will be 1000+% more likely to see and respond to your interest. My bio is here: http://linkedin.com/in/abhaykagarwal/

Our entire interview process is extensively documented here. The interview process is going to look roughly like the following:

1. An introductory call with a founding team member.
2. You will talk to another team member, who will want to know more about your background, prior projects, and hard/interesting problems that you have worked on. We'll likely be basically grilling you on specifics, and this interview might come off as "insistent", or "demanding". This is only because we are essentially trying to dig in deep enough to where we're satisfied that you are deeply aware of the technical trade-offs of a given system.
3. We'll have you do a take-home project or exercise (most of you will come here because of that). You'll have essentially unlimited time to work on this, but the longer that you work on it, the higher the standard becomes. On the other hand, working more on this also shows higher passion and conviction. Overall, we are extremely comfortable with both the "ninja that did this in less than an hour" types, but also the more methodical types. So even if you're in a hurry, give it a shot!
4. We will do a team-wide set of interviews where you will be speaking with everyone on the team. As long as we are small, we will operate on a consensus basis for hires, so we all want to be highly excited to hire you. Please see the "Team Interview" section for more information on how to prepare for this.
5. If we are all very high conviction in your capabilities after all those conversations, the last thing we will do is pay you to fly out to meet us, and work with us for a while (at least a day, less than a week at most). You have agency in this, in terms of when you can come, how long you can stay, and what you want to work on while you're here. We ue you have fun!
6. Finally, we'll make you an offer and/or discuss logistics. In most cases, we'll have already discusses some of the specifics already, so this won't be full of bad surprises.