About Poly

Poly is the most advanced personal cloud storage platform for your files and media. Our platform is designed to be so good, that it'll replace Finder/File Explorer on your desktop. Find anything with multi-modal, natural language search. Browse with lightning fast caching and loading. Convert file types, open in your favorite apps, and tag/rank your content for instant access.

Built on a foundation of incredible research and engineering, Poly is revolutionizing a technology that people use every day by reimagining the stack top to bottom using advanced proprietary AI.

About the role

Skills:Python, GPU Programming, OpenCV, Machine Learning, Torch/PyTorch, Kubernetes

About Us

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

Our file storage platform will "replace Finder and File Explorer" as the de-facto user experience for managing your personal files (we will host them as well). We know, it's an insanely ambitious goal, but we think it's not only possible but necessary to finally do away with these legacy file systems.

Concretely, here's three reasons why people need Poly:

Find anything using natural language and multi-modal queries ("images in the style of X from last week"), so that you never lose track of your stuff again.
Auto-upload/sync content from your GenAI apps via auth or browser plugin, complete with metadata.
Create projects and workspaces where you can pin files, save folders that dynamically update, and mention notes/links for reference. Spend less time managing files and more time getting stuff done.

About You

We are looking for a founding ML expert to join our existing two ML domain experts in building the intelligence behind Poly. You will need to wear many hats, but the only hat you'll never remove is your attention to detail and dedication towards excellence. ou must have significant experience (or significant enthusiasm) in building multi-modal ML systems. Specifically, we would pay attention to your expertise in the following:

Core architectural considerations for 10B+ parameter models, including layer architecture, attention mechanisms, training optimization, evaluation, inference, and tuning.
Extensive knowledge, experience, or enthusiasm for PyTorch-based modeling workflows using Python as a core modeling language. You have authored entire systems from scratch, preferably along with a github or research paper.
Linear algebra and multi-variate statistics, as well as general knowledge of core ML concepts such as gradient descent/backpropagation, AdaGrad/Adam, cross-attention, cross-entropy, and affine transformations.
Multi-modality considerations including contrastive modeling, ViTs, tokenization.
Inference optimizations, such as low-rank fine tuning, distillation, speculative decoding or other similar generation-side enhancements
AI Ops, including model hosting, training pipelines, versioning, experiment tracking, and reproducibility
Data ops, including data loading, synthetic data generation, augmentation, and RLHF/reward modeling.

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/

Technology

We have built an entirely home-grown AI platform and self-trained models for search and ranking. That includes our own inference stack, training pipeline, and so forth. We don't rely on any 3rd-party APIs or services because we want to control our own destiny and build the absolute best tools possible.

Interview Process

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

An introductory call with a founding team member.
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.
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!
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.
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 hope you have fun!
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.

Poly remote jobs

Founding Infrastructure/Cloud Architect

Founding ML Expert

Founding Platform/Full-Stack/Frontend Engineer - Generative AI-Enabled Design Asset Marketplace - Founding team at a YC startup backed by Bloomberg, Felicis, Figma