About this role.
Atomic is hiring a founding forward-deployed AI engineer to build production, industry-specific AI agents for software businesses in its operating portfolio. The role combines hands-on implementation, customer workflow discovery, model selection and optimization, evaluation infrastructure, and migration from legacy systems of record. The engineer is expected to deploy a working agent to a live customer environment within 30 days and create a repeatable AI engineering playbook. This is a highly autonomous, build-heavy position with limited people management responsibility and frequent direct customer engagement.
Role DNA
A quick view of the complexity, pace, ownership and collaboration implied by the job description.
Job Complexity
5/5Pace & Pressure
5/5Autonomy Level
5/5Communication Load
5/5Salary analysis
Estimated compensation compared with the broader US market for similar roles.
Core skills
Skills and capabilities most closely associated with this opportunity.
Sample interview questions
I would explain the workflow baseline, including manual effort, error rates, and turnaround time, then describe the agent architecture, integrations, guardrails, and rollout process. I would quantify results using measures such as task completion rate, human-review rate, accuracy, latency, and cost per completed workflow.
I start with the task requirements, failure costs, data availability, latency, and unit economics. I use deterministic systems for structured and high-confidence logic, prompting for flexible low-data tasks, RAG for grounded knowledge retrieval, fine-tuning for recurring behavior or format gaps, and routing when task complexity justifies different models.
I would first capture representative historical cases and define task-specific acceptance criteria with domain experts. The harness would combine automated checks, labeled golden sets, adversarial cases, regression testing, cost and latency telemetry, and a human-review loop for high-risk failures.
I would profile the source data, identify critical entities and workflow dependencies, establish a canonical schema and validation rules, and migrate incrementally with reconciliation checks. AI can assist in classification and normalization, but I would retain audit trails, confidence thresholds, and human review for ambiguous or business-critical records.
I would spend the first days observing the workflow and selecting a narrow, high-value use case with clear boundaries. I would ship a secure minimum viable integration quickly, instrument it heavily, use human approval where risk warrants it, and iterate from measured customer feedback rather than trying to automate the entire process at once.
We’re a stealth holding company acquiring and operating software businesses in overlooked, mission-critical industries – the systems that run the physical economy but have been ignored by Silicon Valley. We’re rebuilding them as AI-native from the inside out.
We’re small, well-capitalized, and moving fast. This is not a research lab. This is a build shop with real customers, real revenue, and real P&Ls.
The Role
You will be the AI engineer. You’ll own:
Building the industry-specific GPT: the reasoning layer that encodes the domain expertise of each vertical we operate in.
Embedding AI-native functionality directly into our AI-native system of record: not bolt-on chatbots. Core workflow replacement. Agents that do the job.
Moving legacy systems of record datasets into new ai-native ones: Core workflow replacement. Agents that do the job.
Making it repeatable: Every new company we bring into the portfolio should benefit from the playbook, model stack, eval harness, and infra you build.
Managing AI talent across our operating companies: you’ll be a player-coach for the AI engineers embedded at each sub.
Split: 90% building. 10% managing. If that ratio scares you, this isn’t your role. If it excites you, keep reading.
What You’ll Do…
Ship a working domain-specific agent into a live customer environment in your first 30 days
Build the eval infrastructure that lets us measure whether these agents are actually replacing human workflows (not vibes)
Fine-tune, distill, or route across frontier + open models depending on the economics of each vertical
Sit with customers. Watch them work. Instrument their workflows. Rebuild them.
Write the AI engineering playbook that every future portfolio company inherits on day one
Who You Are…
You are a builder first. You’ve shipped things people actually use. Alone or as the clear technical lead.
You are a forward-deployed engineer. You’ll fly to a customer site, sit in their office, and understand the business before you write a line of code.
You’ve been called a 10x engineer and it wasn’t in a self-review.
You have strong opinions about when to fine-tune vs. prompt vs. RAG vs. agent vs harness and you can defend them with numbers.
You’ve built with the current frontier stack (Claude, GPT, Gemini, open models) in production, not just in a Jupyter notebook.
You are allergic to slide decks and comfortable with ambiguity, small teams, and unglamorous industries.
What This Role Is Not…
A research role. We ship.
A management role. You build.
A “Head of AI” title for LinkedIn. You will be in the codebase daily.
A place for people who need process, roadmaps handed to them, or a 20-person team to feel productive.
We are focused on building a diverse and inclusive workforce. If you’re excited about this role, but do not meet 100% of the qualifications listed above, we encourage you to apply.
—–
Atomic is an Equal Opportunity Employer and considers applicants for employment without regard to race, color, religion, sex, orientation, national origin, age, disability, genetics or any other basis forbidden under federal, state, or local law.
Please review our CCPA policies here.
Upload your resume now
- To unlock remote work opportunities and be discovered by global employers.
This job listing has been manually reviewed by the Jobicy Trust & Safety Team for compliance with our posting guidelines, including verification of the company's legitimacy, accuracy of job details, clarity of remote work policy, and absence of misleading or fraudulent content.
