SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: USD 175,000 - 275,000 / annual
Etched is building custom hardware and software infrastructure for frontier AI inference. The company co-designs chips, racks, software, and manufacturing to optimize throughput and latency for both prefill and decode workloads, backed by hundreds of millions in Series A funding from top-tier investors.
As MTS - Agent Research, you will lead the research direction for self-improving agentic systems that reliably complete difficult, long-running engineering tasks. You'll establish a virtuous loop: better agents accelerate system development and optimization, improved systems expand experimental capacity, and the resulting data further enhances agent capabilities.
You will own the research agenda for agentic systems across Etched, identifying critical bottlenecks and prioritizing promising approaches. Working directly with domain engineers across ASIC development, inference, and supercomputing, you'll diagnose agent failures and translate them into concrete research questions with measurable evaluations.
Key responsibilities include: developing systems for long-running tasks (memory, context management, tool use, planning, multi-agent coordination); building evaluations that measure correctness, reliability, and efficiency on real engineering tasks; training custom models using SFT, RL, and distillation with the agent harness in the training loop; and scaling experiments to hundreds or thousands of concurrent agents while maintaining reproducibility and cost control.
This is a hands-on research leadership role requiring strong research judgment about where improvements in memory, context, tooling, or structured workflows unlock new capabilities. You'll set technical direction, write and debug code, and carry ideas through to production. Success is measured by demonstrable improvements in work Etched can accomplish with disciplined compute usage.
The role is fully in-person at the San Jose office (Santana Row). Etched operates without boundaries between engineering and research, expecting all technical staff to contribute across disciplines.
REQUIREMENTS (Must-have):
- Strong research taste with a record of turning ambiguous problems into clear hypotheses, decisive experiments, and useful systems
- Excellent engineering ability; ability to move between research exploration, agent experimentation, low-level debugging, and production execution
- Experience building agents for complex, multi-step tasks with practical understanding of context, memory, tools, evaluation, and failure recovery
- Hands-on experience with LLM post-training, including SFT and RL
- Track record of making meaningful progress under compute and data constraints; focus on experiment quality and cost efficiency
- High agency and comfort with unfamiliar domains; ability to find problems, acquire context, and drive work forward
- Ability to set technical direction and work closely with domain experts to turn research into trusted, usable systems
NICE-TO-HAVE:
- Experience with ASIC design or verification, accelerator kernels, inference systems, or large-scale compute infrastructure
