Mission

Use deep clinical expertise and technical fluency to shape and safeguard Pathos’ AI Therapy. Our product exists to create maximal therapeutic impact. We are comfortable with clients leaving the app once their problems are resolved. Engagement and retention matter only to the extent that people cannot be helped if they do not engage.

We focus on making our product and AI accessible so that people who currently receive little or no support can actually get help, including at global scale. We care deeply about safety, especially in our interactions with vulnerable populations such as people in crisis states.

Your mission at Pathos therefore is to help us maximize safety, accessibility, and therapeutic benefit. You will be a critical team member working cross-functionally with multiple teams. You will influence how we build the product across onboarding, the interventions we offer, and how we track outcomes, and you will play a core role in developing our AI.

If your work here succeeds, you will help save millions of lives.

Outcomes

1. Improve quality of AI therapy. Whether through prompt/context engineering, labeling/creating training data, analysis and insights, or collaboration with the AI research team, you will help us improve the effectiveness of our AI therapy.

2. Safety and guardrails. You will own safety of our AI. You will identify potentially problematic in model behavior, improve monitoring, improve guardrails and AI policies, and ensure that our AI is not only maximally effective but also safe.

3. Clinically grounded evals and data. You will work with the AI research team to iterate and improve on how we evaluate our AI before and after deployed into production.

4. Better training data. You’ll work with the AI research team to improve the data quality used to improve our models.

5. Context engineering and prompts.Work in the codebase to improve prompts and context flows to directly improve the quality and safety of the AI

6. Product and UX influence. Work with the product team to identify changes to the product that can improve the effectiveness our AI, improve how we measure outcomes, and improve the overall impact of of using our product, which goes beyond the AI alone.

7. Scientific and data-informed work.

   Formulate practical hypotheses, participate in experiments and analyses, and use data to guide recommendations and decisions.

Competencies

1. Clinical expertise and licensure. Experienced therapist with strong grounding in evidence-based care and an active clinical license or clear equivalent. Comfortable making nuanced judgments about quality, risk, and appropriateness and capable of making nuanced guidance on how to improve our AI.

2. Research and scientific mindset. Trained to think in hypotheses and tests, not just intuition. Familiar with research methods and basic statistics and able to help design and interpret simple experiments and analyses.

3. LLM and AI literacy. Understands how LLMs behave in practice, common failure modes, and the basics of prompt and context design. Curious about AI capabilities and limitations and can translate them into clinical and product implications.

4. Writing. Writes fast, clearly, concretely. Can write effective prompts and new training data. Explains clinical ideas in plain language to teammates.

5. Product and business awareness. Understands that the goal is to maximize therapeutic benefit at scale in a real startup. Is able to prioritize accordingly

6. High throughput execution. Comfortable moving quickly with incomplete information, doing both tedious labeling and ambiguous exploration. Relentless. Fast. Focused.

7. Collaboration and influence. Works well with researchers, engineers, PMs, designers, and leadership. Listens, incorporates constraints, and can persuade through examples and clear reasoning while staying open to changing course.

Great to Have

Technical fluency. Our ideal candidate would be comfortable working in an IDE, using github, internal tools, etc. You do not need to be an engineer or to build systems but you should be able to edit prompts, configs, and docs without being blocked.

Basic SQL. Is able to write basic SQL commands or is willing to learn.