Job DescriptionJob DescriptionAbout the Company\nOur client is a pre-seed AI startup in San Francisco building AI coworkers for IT teams: a security-focused product where an agent registers as a governed identity in a customer's directory, requests scoped access per task, escalates for human approval, and drives systems it was never given an API for.\nAbout the Role\nThis role builds the layer that turns model capability into systems that actually work for users. You'll develop the core agent harness (the execution loop, tool-use strategies, context construction, and model-facing experimentation) and iterate on agent behaviors across real customer workflows and long-horizon tasks. A defining stance: the creative step happens once, at authoring, and what executes afterward is deterministic, compiled, type-checked code rather than stochastic tool-chaining. You'll own that boundary between what the model decides at runtime and what ships as code.\nYou'll build and run evals against replicas of real customer environments rather than synthetic tasks, reliable enough to gate a release; analyze production failures and attribute them to the right layer (model, prompt, tool contract, environment state, or retry logic); and extend the computer-use agent to systems with no API, owning the guarantees that let it run in production at all: a fresh microVM per task, credentials injected at session start and never seen by the model, every action recorded, and any session killable mid-run.\nWhat You'll Own\n\nThe core agent harness: execution loop, tool-use strategies, context construction, and model-facing experimentation\nAgent behaviors across real customer workflows and long-horizon tasks\nThe boundary between runtime model decisions and deterministic, compiled, type-checked execution\nEvals against replicas of real customer environments, reliable enough to gate releases\nProduction failure analysis and systematic robustness improvements, attributed by layer\nExtending computer-use to API-less systems, with production guarantees (per-task microVM, model-blind credentials, full action recording, killable sessions)\nFeedback loops and data systems that get better real-task data into eval and training\n\nRequirements\nMust-Have\n\n4+ years of engineering experience\nExperience with agent frameworks or tool-using LLM systems\nStrong in Python or TypeScript, comfortable with modern AI tooling\nExperience with model evaluation, fine-tuning, or prompt design\nCan own systems end-to-end and debug across the stack\nThinks in terms of systems and user outcomes, not just model metrics\nEnjoys debugging messy, real-world failures and turning them into improvements\nWants to work in the layer that turns model capability into systems that work for users\nAble to work in person in San Francisco, Monday to Friday, at startup intensity\n\nStrong candidates may also\n\nHave built computer-use or browser-automation agents\nHave experience with virtualization and sandboxed execution environments, and scaling them\nHave done AI research and published at top conferences\nHave shipped something where correctness had to survive partial failure (idempotency, resumability, compensating actions)\nHave integrated enterprise SaaS APIs (identity providers, directory services, ticketing, ERPs)\nHave experience with large, messy datasets or production logs\nHave been an early engineer somewhere they also had to talk to customers\n\nKey Details\n\nLocation: San Francisco, CA (in-person, Monday to Friday)\nEmployment: Full-time\nLevel: Senior / Staff\nExperience: 4+ years\nSalary: $200,000 – $300,000 USD per year\nVisa Sponsorship: Available (H-1B, O-1, OPT)\n\nBenefits & Perks\n\nMeals in office\nHealth insurance\nUnlimited PTO\n\nInterview Process\n\nCulture screen\nSecond culture screen\nTechnical interview\nOne-day onsite\nOffer\n
Senior/Staff Applied AI Engineer, Agent Harness
Apply
Click on the below icons to share this job to Linkedin, Twitter!
Job DescriptionJob DescriptionAbout the Company\nOur client is a pre-seed AI startup in San Francisco building AI coworkers for IT teams: a security-focused product where an agent registers as a governed identity in a customer's directory, requests scoped access per task, escalates for human approval, and drives systems it was never given an API for.\nAbout the Role\nThis role builds the layer that turns model capability into systems that actually work for users. You'll develop the core agent harness (the execution loop, tool-use strategies, context construction, and model-facing experimentation) and iterate on agent behaviors across real customer workflows and long-horizon tasks. A defining stance: the creative step happens once, at authoring, and what executes afterward is deterministic, compiled, type-checked code rather than stochastic tool-chaining. You'll own that boundary between what the model decides at runtime and what ships as code.\nYou'll build and run evals against replicas of real customer environments rather than synthetic tasks, reliable enough to gate a release; analyze production failures and attribute them to the right layer (model, prompt, tool contract, environment state, or retry logic); and extend the computer-use agent to systems with no API, owning the guarantees that let it run in production at all: a fresh microVM per task, credentials injected at session start and never seen by the model, every action recorded, and any session killable mid-run.\nWhat You'll Own\n\nThe core agent harness: execution loop, tool-use strategies, context construction, and model-facing experimentation\nAgent behaviors across real customer workflows and long-horizon tasks\nThe boundary between runtime model decisions and deterministic, compiled, type-checked execution\nEvals against replicas of real customer environments, reliable enough to gate releases\nProduction failure analysis and systematic robustness improvements, attributed by layer\nExtending computer-use to API-less systems, with production guarantees (per-task microVM, model-blind credentials, full action recording, killable sessions)\nFeedback loops and data systems that get better real-task data into eval and training\n\nRequirements\nMust-Have\n\n4+ years of engineering experience\nExperience with agent frameworks or tool-using LLM systems\nStrong in Python or TypeScript, comfortable with modern AI tooling\nExperience with model evaluation, fine-tuning, or prompt design\nCan own systems end-to-end and debug across the stack\nThinks in terms of systems and user outcomes, not just model metrics\nEnjoys debugging messy, real-world failures and turning them into improvements\nWants to work in the layer that turns model capability into systems that work for users\nAble to work in person in San Francisco, Monday to Friday, at startup intensity\n\nStrong candidates may also\n\nHave built computer-use or browser-automation agents\nHave experience with virtualization and sandboxed execution environments, and scaling them\nHave done AI research and published at top conferences\nHave shipped something where correctness had to survive partial failure (idempotency, resumability, compensating actions)\nHave integrated enterprise SaaS APIs (identity providers, directory services, ticketing, ERPs)\nHave experience with large, messy datasets or production logs\nHave been an early engineer somewhere they also had to talk to customers\n\nKey Details\n\nLocation: San Francisco, CA (in-person, Monday to Friday)\nEmployment: Full-time\nLevel: Senior / Staff\nExperience: 4+ years\nSalary: $200,000 – $300,000 USD per year\nVisa Sponsorship: Available (H-1B, O-1, OPT)\n\nBenefits & Perks\n\nMeals in office\nHealth insurance\nUnlimited PTO\n\nInterview Process\n\nCulture screen\nSecond culture screen\nTechnical interview\nOne-day onsite\nOffer\n
yes
Send Us Your Feedback
Sign
In
Please check your email. We have sent you a password reset link.
This link will expire in one hour.
Please input your account's email
Apply to Senior/Staff Applied AI Engineer, Agent Harness
David Joseph & Company • San Francisco, California
✓
Application Started!
Redirecting you to complete your application...
Get $5 added to your wallet for registering an account. Refer others and get $1 for each successful referral. Earn while you search for jobs and redeem in gift cards!Register to search from millions of jobs across the United States. Upload your resume and get matched with top employers today!
Welcome to SnapRecruit! Your account has been created via Google.
Get $5.00 Added to Your Wallet!
Upload your resume below and apply to this job to have $5.00 added directly to your wallet, activate your account, and qualify for high-priority employer matches.
Perks & Rewards You Receive with SnapRecruit:
$5.00 Resume Bonus
Added automatically to your wallet when you apply with your resume.
$1.00 Per Referral
Earn $1.00 cash rewards for every friend who joins SnapRecruit in the US.
Earn While You Search
Get credited $0.02 - $0.05 every time you explore sponsored job listings.
Amazon eGift Cards
Redeem your earned wallet funds for Amazon eGift Cards delivered to your email.
Get $5 added to your wallet for registering anRegister to search from millions of jobs across the United States. Upload your resume and get matched with top employers today!
account.
Refers others and get $1 for each successful referral.
Earn while you search for jobs and redeem in
gift cards!
Upload your resume for a free professional assessment
Loading, Please wait..!!
Send this job to my email
Please wait..!!
Confirmation
Are you sure?
List of resumes
Confirmation
It is working for Chrome Mozilla only for now
We use essential cookies to keep SnapRecruit secure and reliable. With your permission, we also use optional analytics, AI support chat, and push-alert cookies to improve your job search experience. You can change or withdraw your choice anytime.
Privacy Policy •
Cookie Policy •
Your Privacy Rights