AI Researcher

Traversal•New York, NY

4h•Onsite

About The Position

About Traversal Traversal is the AI Site Reliability Engineer (SRE) for the enterprise—already trusted by some of the largest companies in the world to troubleshoot, remediate, and even prevent the most complex production incidents. Our mission is to free engineers from endless firefighting and enable them to focus on creative, high-impact work. Our roots remain deeply embedded in AI research, and we’re channeling that scientific rigor and creativity into building the premier AI agent lab for the enterprise. Hence, what we’re proudest of is assembling the most talented yet nicest group of individuals, including researchers from MIT, Harvard, and Berkeley, to world-class engineers from industry: Citadel Securities, Cockroach Labs, Datadog, DE Shaw, ServiceNow, Glean, Perplexity, Pinecone, and more, to take on one of the hardest problems for AI to solve. Without the entire team, none of this would be possible. The Role As an AI Researcher at Traversal, you'll work on improving the accuracy and speed of our agents — systems that autonomously diagnose and resolve production incidents for some of the world’s largest enterprises. This is a hands-on, production-oriented research role. You will design experiments, run them on real data, and ship improvements. Not flag them — ship them. You'll work end-to-end: identify a failure mode in agent reasoning, design an intervention, evaluate it against real customer traces, and get it into production. The tooling and infrastructure are in place. The research problem is hard. The feedback loop is fast. This is not a publish-papers role. It is a make-the-agent-work role – build-and-ship cutting edge AI.

Requirements

PhD in Computer Science, Electrical Engineering, Statistics, or a related technical field; demonstrated depth in LLMs, agents, or applied machine learning
Deep applied AI expertise, including strong working knowledge of LLMs, transformers, reinforcement learning, or neural networks in agentic systems
Strong judgment in model evaluation and experimental iteration to improve product accuracy and behavior
Strong software engineering depth, with the ability to work effectively in a complex production codebase and ship production-quality code
Some experience shipping AI or ML systems to production
Ability to run rigorous experiments, interpret results, and quickly translate learnings into product improvements
Startup or early-team experience, with comfort operating in ambiguous environments and building without mature infrastructure

Nice To Haves

Experience in SRE, observability, or backend systems, especially when paired with strong AI/ML depth
Experience with RLHF, synthetic data pipelines, or LLM evaluation tooling
Contributions to open-source agent frameworks such as LangGraph, DSPy, or similar
Research experience in LLMs, agents, or reinforcement learning, including publications in venues such as NeurIPS, ICML, or ICLR; top-tier conference publications are a plus

Responsibilities

Prototype and evaluate prompting strategies, reasoning workflows, and tool-use policies for agents operating on large-scale observability data and complex troubleshooting workflows.
Ship improvements to production.
Build and maintain eval harnesses that measure real accuracy improvements on actual customer incident types — not just benchmark scores.
Own the loop from hypothesis to production measurement.
Work closely with AI engineers, infrastructure teams, and product leads to bring research into production and close the loop between experimentation and impact.
Stay on the Frontier: Track developments in LLMs, agent architectures, and AI alignment, translating insights into actionable improvements for Traversal’s domain.
Apply fine-tuning, reinforcement learning, and reward modeling techniques to align AI behavior with real-world SRE workflows.
Design pipelines to generate synthetic incidents and observability signals, enabling scalable training and testing in data-scarce environments.