Research Engineer

Console•San Francisco, CA

19h

About The Position

Console is building the AI agents that power autonomous enterprises. We use AI to autonomously resolve IT, HR, Legal, Finance, Security, and Ops tasks the moment they come in from Slack or Teams. We're building toward a future where enterprise operations are fully autonomous; with no tickets, approvals, or manual processes slowing companies down. Our platform enables companies like Databricks, Figma, and Cursor to delegate entire categories of work to AI agents that understands their org, takes action across systems, and gets smarter over time. We've raised $29M, backed by Thrive Capital (investors in OpenAI, Stripe, GitHub) and a small group of founders who have built category-defining companies. We’re unusually selective about the people we work with. We hire builders who span functions, optimize for speed, and care deeply about their craft. We do not care about pedigree. We have a lot of work ahead of us. Join us. About the role We’re hiring a Research Engineer to push Console’s core agent loop from powerful to truly self-improving. As more customers rely on Console to automate critical back-office operations, we need engineers who can turn real production traffic into better agents, evals, and specialist models. This role sits at the intersection of applied AI engineering and research. You’ll build the systems that let us measure, debug, and improve agent behavior in real-world conditions: production traces, offline replay, labeling workflows, eval harnesses, prompt and program optimization, and fine-tuning loops for high-volume agent tasks. One of your first major focus areas will be improving how Console’s agents reason over complex enterprise context. Our agents need to understand users, apps, devices, tickets, licenses, policies, and customer-specific data well enough to answer questions and take action reliably. You’ll help turn this into a compounding optimization loop: using production traces, evals, assisted labeling, prompt/program optimization, and targeted model adaptation to make the system measurably better over time. You’ll work closely with product, engineering, and leadership to ship improvements into production quickly, while helping define what research at Console looks like as we scale.

Requirements

Product-minded research engineer who enjoys building real systems people depend on.
Strong technical background in software engineering, machine learning, or applied AI, demonstrated through an advanced degree and/or equivalent experience building production AI systems.
Strong software engineering fundamentals and good judgment for designing, building, and debugging complex systems.
Experience building evals for AI systems, including datasets, judges, metrics, offline replay, tracing, or regression testing.
Practical understanding of modern model adaptation and post-training methods, including LoRA/QLoRA, SFT, distillation, preference optimization, reward modeling, DPO/GRPO, and reinforcement learning from verifiable feedback.
Ownership mindset: you drive projects end-to-end, move quickly from real usage, and care about shipping measurable improvements.
Enjoy following SOTA research into new models, agent architectures, evals, post-training methods, and optimization techniques.

Responsibilities

Build and improve AI systems that operate in real-world conditions, where reliability, speed, cost, and adaptability matter.
Build the eval and optimization loop for Console’s core agents, turning real production usage into measurable improvements in quality, latency, and reliability.
Systematically improve agent behavior across prompts, programs, routing logic, constraints, and model adaptations, applying techniques like DSPy and GEPA where useful.
Fine-tune, adapt, and evaluate specialist models for repeatable, high-volume agent tasks where Console has clear production feedback or verifiable quality signals.
Work across the stack when needed, from traces and eval infrastructure to agent orchestration, product workflows, and customer-facing AI behavior.