Strategic Projects Intern

Job highlights Identified by Google from the original job post Qualifications • Have 1+ first-author publications at top ML conferences (NeurIPS, ICLR, ICML, etc.) • Have trained or substantially contributed to training 1B+ parameter LLMs or VLMs • Come from a strong research background (MS or PhD) in areas like: 1) Computer-Use Agents, 2) Vision-Language Models, 3) Computer Vision, 4) Robotics, 5) Reinforcement Learning • Are deeply fluent in PyTorch, HuggingFace, or similar frameworks • Have experience with modern RL training frameworks (e.g • Atropos, NeMo-RL, VERL, etc.) • Are high-agency, self-directed, and comfortable owning ambiguous research problems • Want to work in person in San Francisco with a small, fast-moving team • Are excited by healthcare as a real-world application (prior domain experience not required) • Initial conversation via CoffeeSpace • Research deep dives + technical discussions with the team • 8 more items(s) Benefits • Compensation: $175k–$250k base + meaningful equity • Partnering directly with healthcare operators to anchor research in economically meaningful tasks • Competitive base + meaningful equity • Unlimited PTO • Comprehensive health, dental, and vision coverage • 401(k) • Free lunch + dinner daily • 4 more items(s) Responsibilities • Training and evaluating computer-use agents that operate across complex, messy interfaces • Advancing model capabilities through post-training, RL, and rigorous evaluation, not just scaling • The goal: push state-of-the-art model capabilities while staying tightly coupled to real deployment constraints • As a Member of Technical Staff (AI Researcher), you’ll help drive our core research agenda — from identifying capability gaps in current models to building the environments, data, and training pipelines needed to close them • This is a research-heavy, high-ownership role with real influence over what the team works on, publishes, and ships • You’ll work closely with researchers, engineers, and healthcare partners — owning problems end-to-end rather than contributing to narrow slices of a larger org • Design novel evals, benchmarks, and RL environments that reflect real healthcare workflows • Develop, train, and evaluate computer-use agents operating in complex interfaces • Build and maintain data pipelines that transform raw human data into high-quality training and eval assets • Post-train large models (LLMs / VLMs) to advance real-world capabilities • Write and publish research papers at top-tier ML venues • Work across the stack — models, tooling, infra, internal workflows, and product surfaces • On-site interviews and collaborative problem sessions • 10 more items(s) More job highlights Job description About the job

This role is being recruited by CoffeeSpace on behalf of Stealth Startup an early-stage AI research startup based in San Francisco working at the intersection of computer-use agents, human data, and healthcare.

We’re identifying a small number of exceptional AI researchers from our network who meet the bar for this role.

If there’s a strong fit, you’ll begin with an initial convers...ation via CoffeeSpace before moving into direct conversations with the founding and research team.

Member of Technical Staff (AI Researcher)

Location: San Francisco (on-site)

Compensation: $175k–$250k base + meaningful equity

Employment type: Full-time

Visa support: Case-by-case (strong candidates encouraged to apply)

What we are building (Frontier AI × Real-World Healthcare)

Most frontier models are evaluated on synthetic or narrow benchmarks that poorly reflect how AI is actually used in the real world — especially in healthcare.

We are changing that by:

• Building novel datasets, environments, and benchmarks grounded in real healthcare workflows

• Training and evaluating computer-use agents that operate across complex, messy interfaces

• Advancing model capabilities through post-training, RL, and rigorous evaluation, not just scaling

• Partnering directly with healthcare operators to anchor research in economically meaningful tasks

The goal: push state-of-the-art model capabilities while staying tightly coupled to real deployment constraints.

The role

As a Member of Technical Staff (AI Researcher), you’ll help drive our core research agenda — from identifying capability gaps in current models to building the environments, data, and training pipelines needed to close them.

This is a research-heavy, high-ownership role with real influence over what the team works on, publishes, and ships.

You’ll work closely with researchers, engineers, and healthcare partners — owning problems end-to-end rather than contributing to narrow slices of a larger org.

What you’ll do

• Design novel evals, benchmarks, and RL environments that reflect real healthcare workflows

• Develop, train, and evaluate computer-use agents operating in complex interfaces

• Build and maintain data pipelines that transform raw human data into high-quality training and eval assets

• Post-train large models (LLMs / VLMs) to advance real-world capabilities

• Write and publish research papers at top-tier ML venues

• Work across the stack — models, tooling, infra, internal workflows, and product surfaces

You’re likely a strong fit if you:

• Have 1+ first-author publications at top ML conferences (NeurIPS, ICLR, ICML, etc.)

• Have trained or substantially contributed to training 1B+ parameter LLMs or VLMs

• Come from a strong research background (MS or PhD) in areas like: 1) Computer-Use Agents, 2) Vision-Language Models, 3) Computer Vision, 4) Robotics, 5) Reinforcement Learning

• Are deeply fluent in PyTorch, HuggingFace, or similar frameworks

• Have experience with modern RL training frameworks (e.g. Atropos, NeMo-RL, VERL, etc.)

• Are high-agency, self-directed, and comfortable owning ambiguous research problems

• Want to work in person in San Francisco with a small, fast-moving team

• Are excited by healthcare as a real-world application (prior domain experience not required)

Why us?

• Founding-level impact at a frontier AI research company

• Deep intersection of cutting-edge AI + real deployment constraints

• Backed by Tier-1 investors, founded out of Stanford

• Small, talent-dense team with real ownership and velocity

Benefits include:

• Competitive base + meaningful equity

• Unlimited PTO

• Comprehensive health, dental, and vision coverage

• 401(k)

• Free lunch + dinner daily

Interview process

• Initial conversation via CoffeeSpace

• Research deep dives + technical discussions with the team

• On-site interviews and collaborative problem sessions

Next steps

• Apply via this LinkedIn job post

• We’ll review your profile and reach out if there’s a strong match

• If aligned, we’ll introduce you directly to the founders

If this role isn’t the right fit, we may suggest other high-signal AI and early-stage startup roles we’re recruiting for — always with your permission.

A quick note on authenticity

This is a real, active role that CoffeeSpace is recruiting for on our behalf.

We don’t post speculative roles, and we only work with companies where we have direct access to the hiring team. Show full description Report this listing Loading...