Buy 'yes', help me find something I'm happy with, earn mana. Or just trade your beliefs. A low implied probability in this market will motivate me to direct my efforts toward other goals.

I'm looking for a job doing ML (machine learning) engineering. This market resolves 'yes' if I have a firm agreement on a position that matches. A "firm agreement" means I've signed an offer letter or at least received a formal detailed offer and accepted it. I don't have to have actually started the job. A job counts as ML engineering if it's primarily about software engineering for machine learning systems. Titles such as "Machine Learning Engineer", "Research Engineer", "Member of Technical Staff" at an AI company, etc count. A software engineering job that doesn't involve ML or involves ML for less than 30% of its duties doesn't count.

Market deadline is midnight April 30th eastern time. If there's no firm agreement on a matching job this market resolves 'no'.

As for my background, I've had two software engineering jobs, neither of which involved ML. In 2017-2018 I worked on the backend of a programmable text messaging platform. And in 2018-2020 I worked at a company developing a new cryptocurrency. In January 2020 I quit that job. I was dissatisfied - I felt like the work I was doing didn't matter. My plan was to take some time off and figure out what I wanted to do next. That lasted a long time, the cryptocurrency launched and my deferred comp meant I didn't have to worry about money for a long time. Now it's a bit more than 4 years later and I do. Recently I've been working on an ML project, partly to build legible skills and partly to answer a research question I'm curious about. Here's the copypasta summary I wrote:

I've been trying to build a text-to-image model that is trained without any text labels, using unlabeled images and CLIP for the link between captions and images. Results are promising but inconclusive so far. I think this work is the best representation of what I'm capable of. In the process, I gathered a dataset of 33 million images for training data, including removing redundant images, deduplicating, and taking stills from any videos. I ported a VQGAN implementation from PyTorch to JAX, built an efficient preprocessing pipeline, built transformer models in JAX, and designed and trained baseline models and more sophisticated ones. To support the approach I eventually settled on, I designed and implemented an efficient algorithm to sample unit vectors from a finite set, conditioned on the vectors being inside a spherical cap. For that I needed to write a Python library in Rust to help with constructing the space partitioning data structure used for sampling. The sampling algorithm gets used to generate training examples and the model learns to sample images conditioned on the image's CLIP embedding being within an arbitrary spherical cap.

Here are two blog posts about that project: http://www.echonolan.net/posts/2024-02-21-25m-imgur-images-dataset.html & http://www.echonolan.net/posts/2024-03-09-is-it-possible-to-train-a-text-to-image-model-without-any-text.html and my resume is here http://www.echonolan.net/resume/cv.html.

As for connections I have a friend at Anthropic and a friend at OpenAI. I'm unlikely to apply to OpenAI due to x-risk concerns. I will probably formally apply to Anthropic in the next few days. I also have a friend pretty well hooked into the startup space. And a bunch of other friends who are SWEs that don't have any connections to ML stuff.

I won't trade in this market.