Collaborate with AVL on cutting-edge AI research.

Why this exists

AVL builds, markets, and scales AI products. That work keeps running into open research questions, not abstract ones but questions that decide whether a product works: how to measure whether an AI answer engine cites a brand, how an agent should plan and recover when a step fails, how a market of real people actually adopts a new tool. We would rather study those questions properly than guess at them.

The Research Fellow program is where that study happens. Fellows work with AVL on applied and frontier problems, close to products in market, with real data and a real reason for the answer to be right. Some of it is publishable. All of it is meant to be used.

This is not a lab detached from the world, and it is not unpaid product work dressed up as research. It is a place to do serious work on hard AI problems with people who ship, and to see your findings change what gets built.

What you would work on

Real directions, named honestly. Most fellows go deep on one and keep an eye on the others.

• AI distribution and intent measurement. How to measure whether AI answer engines surface a product, what moves that presence, and how buyer intent shows up across models. This is the research behind Searchalong.
• Agentic systems. How agents plan, use tools, recover from failure, and stay reliable enough to trust with real work. Evaluation and architecture, not demos.
• Computational social science. How real communities and markets adopt new tools, using data at scale to understand behavior instead of guessing at it.
• Applied ML. Retrieval, evaluation, fine-tuning, and the unglamorous engineering that decides whether a model is actually useful in a product.
• Evaluation and measurement. Honest benchmarks and metrics for systems that are hard to measure, so a claim about a model or an agent means something.

If your strongest area is not on this list but sits near it, tell us. We care about the quality of the work, not the label on it.

What you get from it

• Real problems with real data. You work on questions that come from products in market, with the data and context to answer them properly.
• Work that gets used. Your findings feed directly into what AVL builds. You see research turn into a shipped product, not a slide deck that gets filed away.
• Room to publish. When the work is worth writing up, we support publishing it under your name. Good research should travel.
• A network of builders and researchers. You work next to engineers and founders who ship, and other fellows worth knowing. Those relationships compound.
• Direct collaboration, not a queue. You work with the people making the calls, not through three layers of process. Your read on a problem lands.
• Flexible and remote. Set the depth and cadence that fits your other commitments. This is built to run alongside a degree, a role, or independent work.

Who we're looking for

Researchers and technically strong builders who can do rigorous work and still care whether it ships. The specifics vary. We want:

• Graduate researchers and PhD students in ML, NLP, computational social science, HCI, or an adjacent field, who want their work to touch the real world.
• Research engineers who can turn a hypothesis into an experiment and an experiment into a system that runs.
• Strong independent builders without a formal research title who do serious, curious work and can show it.
• Applied ML practitioners who know the difference between a benchmark number and a model that is actually useful.

The bar is not a title. It is rigor, curiosity, and the honesty to report what the data actually shows, even when it is inconvenient. If that is you, we want to hear from you.

How it works

Questions people ask