---
title: 'Research Fellowship - Mechanistic Interpretability at Vmax'
canonical: 'https://feeny.ai/job/research-fellowship-mechanistic-interpretability-vmax-san-francisco-3wdzvbm4d8bc'
type: 'job'
last_seen: '2026-09-10'
---

# Research Fellowship - Mechanistic Interpretability at Vmax

- **Company:** Vmax
- **Location:** San Francisco, CA
- **Posted:** 2026-05-20
- **Last confirmed live:** 2026-09-10
- **Apply:** https://job-boards.greenhouse.io/vmax/jobs/4253430009

## Job description

## About Vmax

Vmax is an applied research lab developing AI capable of open-ended learning. We are building systems to exceed humans in all capacities by optimizing beyond the local maxima of learning from human expertise.

## About the role

LLMs are fantastically powerful and there is a rapidly growing corpus of work devoted to understanding their internal representations and computations. We use the tools of mechanistic interpretability to enhance reinforcement learning by generating intrinsic rewards as a supplement or alternative to downstream human-generated verifiers.

This 3 to 6 month fellowship is for PhD students or equivalent early-career researchers who want to work at the intersection of mechanistic interpretability and reinforcement learning. You will own a focused research project, work closely with Vmax technical staff, and contribute to research publications.

## Responsibilities

- Develop mechanistic interpretability methods for understanding internal representations, features, circuits, and computations in language models and agents.
- Investigate how model internals can be used to generate intrinsic rewards, auxiliary objectives, diagnostics, or training signals for reinforcement learning.
- Design and run experiments that test whether interpretability-derived signals improve learning, exploration, generalization, robustness, or sample efficiency.
- Compare internally derived rewards against baselines such as human-generated verifiers, reward models, task-level outcome rewards, and standard intrinsic motivation methods.
- Use techniques such as probing, activation analysis, sparse autoencoders, causal interventions, feature attribution, or representation analysis to study model behavior.
- Analyze failure modes, including reward hacking, spurious features, non-causal correlations, objective misspecification, and overfitting to narrow evaluation distributions.
- Build research code, evaluation harnesses, and experimental infrastructure that make results reproducible and useful to the broader team.
- Communicate research progress clearly through written updates, internal presentations, and final project outputs.

## Role Requirements

- Currently enrolled in a PhD program in machine learning, computer science, artificial intelligence, computational neuroscience, mathematics, or a related technical field. Exceptional candidates with equivalent research experience may also be considered.
- Track record of research excellence or strong research promise, demonstrated through publications, preprints, open-source work, technical projects, competitions, or publicly available artifacts.
- Working understanding of reinforcement learning.
- Familiarity with mechanistic interpretability, representation analysis, or empirical methods for understanding neural networks.
- Strong programming ability in Python and experience with at least one major ML framework such as PyTorch or JAX.
- Clear written and verbal communication of technical ideas.

## Nice to have

- Experience with LLM post-training methods
- Familiarity with intrinsic motivation, unsupervised RL, auxiliary objectives, representation learning for RL, or curiosity-driven learning.
- Experience with scalable ML experimentation, distributed training, experiment tracking, or reproducible research infrastructure.
- Interest in turning mechanistic understanding into practical training methods, rather than only analyzing models after training.

## Role specific location policy

- This role is based in our San Francisco office; for exceptional candidates we are willing to consider a hybrid arrangement

## About Vmax

## Company Overview
- **One-liner**: Vmax is an applied research lab developing AI capable of open-ended learning by building reinforcement learning systems that exceed human performance across every domain.
- **Entity Type**: Private (funding stage not disclosed)
- **Headquarters**: San Francisco, California, United States
- **Founded**: Not publicly available
- **Founders**: Augustine Mavor-Parker (Co-Founder, CTO) and Matthew Sargent (Co-Founder, CEO)

## Core Business
- **Primary Industry**: Artificial Intelligence / Reinforcement Learning Research
- **Target Customers**: B2B (enterprise AI teams, research institutions) and the broader AI research community (open-source tools)
- **Mission**: To build systems that exceed humans in all capacities by optimizing beyond the local maxima of learning from human expertise, discovering radically new ways of working rather than simply replacing human labor.

## Products & Services
- **Open-Ended Learning Research**: Core R&D in reinforcement learning algorithms and systems that allow agents to define and optimize their own goals, moving beyond human-curated task distributions.
- **Harbor**: Open-source LLM post-training framework designed for running RL environments and scaling post-training experiments.
- **ARES (Agentic Research and Evaluation Suite)**: Open-source evaluation harness for agentic research, enabling rigorous benchmarking of RL-based agents.
- **Unix-CTF & PopuLoRA**: Procedural environments and co-evolution methods for RL self-play, published as research artifacts.

## Market Standing
- **Valuation/Market Cap**: Not disclosed
- **Key Metric**: Team size of 7 employees (as of mid-2026), with monthly headcount growth of +11.1%
- **Notable Investors/Partners**: No investors publicly listed; alumni from top labs (DeepMind, Meta, The Alan Turing Institute, Mila) have joined the team.
- **Growth Signals**: Active hiring (3 open positions as of July 2026), steady LinkedIn follower growth (+11.8% monthly), recent high-traction hires from Meta and Google DeepMind, and public research releases (May 2026).

## Competitive Advantages
- **Unconstrained Learning Approach**: Vmax avoids the local maxima of human-curated tasks by enabling agents to define their own goals — a true open-ended learning paradigm.
- **Research-to-Engineering Pipeline**: Tight coupling between RL research and production infrastructure (e.g., Harbor framework) allows fast iteration from theory to working systems.
- **Talent Density**: Small team (7 people) composed of specialists from top AI organizations, creating a high-expertise, low-bureaucracy environment.

## Strategic Focus
- **Post-Training Innovation**: Shifting LLM post-training from human-curated distributions to automatically generated, agent-defined tasks.
- **Infrastructure for RL Scaling**: Building reusable, reliable RL training, evaluation, and monitoring systems (Harbor, ARES) to support rapid experimentation.
- **Mechanistic Interpretability for RL**: Exploring how internal model representations can generate intrinsic rewards to improve learning and robustness.
- **Publication & Open-Source**: Regularly releasing research (e.g., PROPEL, Unix-CTF, PopuLoRA) and code to attract talent and establish thought leadership.

## Why Work Here
- **Compensation & Equity**: Salary for MTS roles ranges from $300,000 - $500,000 USD (plus likely equity/benefits).
- **Work Arrangement**: Primarily San Francisco office-based; hybrid considered for exceptional candidates.
- **Culture & Impact**: Small, high-autonomy team with no red tape; engineers own ambiguous projects end-to-end, from problem framing to production.
- **Research Focus**: Direct contribution to cutting-edge RL research with real-world deployment; access to top compute and data.
- **Notable Perks**: Opportunity to publish, attend conferences, and work alongside former DeepMind, Meta, and Alan Turing Institute researchers. Strong emphasis on engineering quality and reproducibility.

## Sources
1. [vmax.ai](https://vmax.ai/)
2. [LinkedIn - Vmax](https://www.linkedin.com/company/vmax-ai)
3. [Greenhouse - MTS Applied RL](https://job-boards.greenhouse.io/vmax/jobs/4056502009)
4. [Greenhouse - Research Fellowship](https://job-boards.greenhouse.io/vmax/jobs/4253430009)
5. [GitHub - VmaxAI](https://github.com/VmaxAI)

## Other roles at Vmax

- [Member of Technical Staff - RL Infrastructure](https://feeny.ai/job/member-of-technical-staff-rl-infrastructure-vmax-san-francisco-qwykrpnbfwgm) — San Francisco, CA
- [Member of Technical Staff - RL Algorithms](https://feeny.ai/job/member-of-technical-staff-rl-algorithms-vmax-san-francisco-fd2386sg75px) — San Francisco, CA
- [Member of Technical Staff - Mechanistic Interpretability](https://feeny.ai/job/member-of-technical-staff-mechanistic-interpretability-vmax-san-francisco-f9q03m76tmdt) — San Francisco, CA
- [Open Application – Exceptional Talent](https://feeny.ai/job/open-application-exceptional-talent-vmax-san-francisco-fj7yd501htx8) — San Francisco, CA
- [Research Fellowship - Open Endedness](https://feeny.ai/job/research-fellowship-open-endedness-vmax-san-francisco-cskvdpnjtjay) — San Francisco, CA
- [Member of Technical Staff - Applied RL](https://feeny.ai/job/member-of-technical-staff-applied-rl-vmax-san-francisco-151f2rpje236) — San Francisco, CA
- [Member of Technical Staff - Open Endedness](https://feeny.ai/job/member-of-technical-staff-open-endedness-vmax-san-francisco-6zfvz7zqa58g) — San Francisco, CA