AI Research Engineer

About this role

We need a researcher who builds. You don't just read Arxiv; you ship it. You are steeped in the nuances of LLMs and agentic AI, bridging the gap between training models and applying them to messy, real-world problems.

You are a pragmatic innovator who knows that "State of the Art" means nothing if it doesn't solve the business problem. You quickly grok academic papers, but can cut through the noise to deliver a working MVP. You obsess over evaluation, not just principles.

What You’ll Do

• Advanced Agentic Systems: Architect and implement complex agent behaviors, leveraging multi-agent hierarchies, multi-trajectory inference, and RL environments to solve open-ended tasks.
• Automated Optimization: Move beyond manual prompt engineering. Build and deploy automated optimization pipelines (like DSPy or custom optimizers) to systematically improve agent performance.
• Rigorous Evaluation: Design and own the evaluation stack. Implement LLM-as-a-judge frameworks and domain-specific metrics to ensure our agents are reliable, safe, and improving over time.
• Applied Research: Translate the latest papers on RAG techniques and cognitive architectures into production-ready code that drives our product forward.

Examples of Every-Day Work

• From Paper to Prototype: Read a new paper on tree-of-thought reasoning in the morning and have a working implementation testing against our benchmarks by the afternoon.
• Optimize the "Brain": Diagnose why an agent is hallucinating in a specific edge case and deploy a targeted fix via context engineering or fine-tuning, validating it with regression tests.
• Architect the Swarm: Design a communication protocol for a hierarchy of agents where a "Manager" agent effectively delegates sub-tasks to specialized "Worker" agents without getting stuck in loops.
• Cut the Fat: Reject a complex research proposal because a simpler, heuristic-based approach solves 90% of the user's problem with 10% of the compute.

What We’re Looking For

• Deep AI Fluency: You are steeped in the latest LLM research. You understand the internals of Transformers, but more importantly, the emergent behaviors of agentic systems.
• Research agility: You can quickly digest papers on RAG, multi-agent collaboration, and RLHF, and judge their applicability to our stack.
• Eval Obsession: You know best practices for evaluation like the back of your hand. You don't trust a prompt until you have the metrics to back it up.
• Product Intuition: You grok business concepts. You propose creative technical solutions that align with commercial goals, not just academic curiosity.
• MVP Mindset: You move fast. You prioritize speed of iteration and learning over theoretical perfection.