Research Engineer - Gauntlet

The Role

Gauntlet is seeking an AI Research Engineer to enhance the development, evaluation, and application of AI agents across the business. We currently have agents, including Aria (our in-house generalist agent), operating in production across research, engineering, and business workflows. While some tasks perform exceptionally well, others yield inconsistent results, and changes to models or tooling can introduce unexpected issues. Currently, much of this feedback is subjective, and we aim to establish a more systematic approach to understanding what works, what doesn't, and why.

A significant part of this role involves building evaluation frameworks to measure agent performance on real tasks. You will collaborate with engineers, quantitative researchers, and other teams to identify agent challenges, translate observations into measurable experiments, and use the results to improve our systems.

You will also directly contribute to our existing agents and internal tooling. Initially, the work will be more engineering-focused as you familiarize yourself with our systems, shipping improvements and integrations for team use. Over time, you will undertake more open-ended research, developing better approaches to evaluation, agent behavior, and infrastructure.

Beyond individual agents, a longer-term goal is to build a shared system of knowledge and reusable skills that our agents can draw from and improve over time. You will play a crucial role in developing this infrastructure and evaluating its effectiveness.

This is an applied research role with a significant engineering component. We are looking for someone who enjoys investigating difficult problems, can design effective experiments, and is comfortable building and shipping the resulting solutions.

About Gauntlet

Gauntlet is building the financial systems of the future. While much of onchain finance focuses on point solutions, we operate across the entire stack to offer best-in-class vault products. We currently serve over $1.7B in client TVL across major fintechs/neobanks, protocols, exchanges, and capital allocators in crypto, and increasingly, traditional asset management. Our team combines traditional finance and crypto-native expertise to deliver durable, sophisticated products for institutional clients moving onchain.

Why This Role is Different

This is an opportunity to work at the intersection of AI, quantitative research, and DeFi, tackling problems where AI advancements can directly impact financial markets and onchain strategies.

You will go beyond building individual agents, helping to define how they are evaluated, how they learn from real-world tasks, and how their capabilities scale across the business. With few established benchmarks in this space, you will have the freedom to explore new approaches and take your ideas from research through to production, working directly with the teams who will use them.

What You'll Do

• Build evaluation frameworks for AI agents, including datasets, sandbox environments, and metrics to measure performance on real-world tasks.
• Design experiments to compare models, agent harnesses, tools, and execution environments, identifying regressions and understanding trade-offs between performance, reliability, and cost.
• Collaborate with teams across Gauntlet to translate qualitative feedback about agent behavior into concrete research questions, experiments, and measurable outcomes.
• Investigate agent failures, diagnose whether problems stem from the model, harness, tooling, or prompting, and develop methods to improve performance.
• Build and improve agent harnesses, internal tools, and integrations, including contributing to Aria and other agents used across the business.
• Help develop shared knowledge and reusable skills across agents, creating systems that can learn from previous work and improve over time.
• Work alongside Infrastructure and Security to ensure agents can operate reliably and safely within our production environment.

What We're Looking For

• Hands-on experience building, evaluating, or improving LLM models or agents, whether through professional work, applied research, open-source contributions, or substantial personal projects.
• Strong Python skills and solid software engineering fundamentals. You should be comfortable writing production-quality code and understand the importance of testing, maintainability, and reliability.
• Experience with applied research, machine learning, or data science, including designing experiments, working with datasets, and drawing conclusions from results.
• A good understanding of how LLM agents work beyond prompting, including tool use, harnesses, context management, and common failure modes.
• The ability to take an open-ended problem, break it down into testable questions, and work independently towards a solution.
• Experience building tools or systems that other people use, with an understanding of how to evaluate their effectiveness beyond technical benchmarks.
• Comfort working across research and engineering, collaborating with different teams, and adapting as priorities change.
• Genuine interest in AI agents and how they can be made more capable, reliable, and useful.

Nice to Have

• Experience building LLM evaluation pipelines, benchmarks, reinforcement learning environments, or model testing infrastructure.
• Contributions to open-source agent frameworks, harnesses, evaluation datasets, or benchmarks.
• Experience with agent frameworks such as Hermes, OpenClaw, or similar tools.
• Experience in applied ML research, model training, fine-tuning, or reinforcement learning.
• An advanced degree or postgraduate research experience in computer science, mathematics, statistics, machine learning, or a related field.
• Experience in financial markets, quantitative research, DeFi, or risk modeling.
• Familiarity with AI security considerations, including access controls, prompt injection, and data protection.

You don't need prior experience in crypto or DeFi. Strong research and engineering fundamentals matter more, and you'll work closely with people across Gauntlet who understand the financial systems and strategies our agents support.

We expect strong candidates to have experience working in a professional engineering or research environment, but we're more interested in what you've built and how you approach problems than a specific number of years on your CV.