About the role:

Provectus is a Premier AWS partner and an Anthropic Strategic Partner at the forefront of applied AI, helping enterprises turn Claude, agentic systems, and their own data into measurable business outcomes through bespoke applications, managed services, and advisory engagements. With offices in North America, LATAM, and EMEA, we partner with clients worldwide and are obsessed with reimagining how they operate and compete.

Our work centers on two verticals — Financial Services & Insurance and Healthcare & Life Sciences — where we deploy five pre-built AI Blueprints: Submission Flow, Portfolio Lens, Asset Flow, Revenue Flow, and Evidence Lens. Each Blueprint rebuilds a critical business process front to back, shipped from working code and tuned to a client's specific book, regulators, and operating posture.

We embed engineers and leaders inside client operations as Forward Deployed Engineers (FDE) and Forward Deployed Executives (FDX) — people who learn the work, ship the system, and own the outcome. Our team holds 100+ AWS certifications, is Claude Code certified, and co-delivers Anthropic's Agentic SDLC program, Cowork Activation, and AI Bluprints.

You will do the customer’s job before you automate it.

Most AI engagements fail the same way: someone gathers requirements, someone writes a PRD, and a team ships a workflow nobody uses. We think the requirements-gathering step is the bug. So we remove it.

A Forward Deployed AI Engineer at Provectus spends the first weeks of an engagement in the operator’s seat — as the underwriter, the analyst, the RCM specialist, the claims clinician, whoever actually does the work we’ve been asked to change. You do the job. You learn the constraints from the inside, the ones nobody writes down. Then you sit at a table with that operator and a Forward Deployed Executive and rebuild the function from first principles — and you are the one who builds it.

Three things define how you work:

Embedded, not engaged. You are part of the customer’s team and inside their process — not a vendor running a project alongside it.

Real tasks, not scope. You are not fenced into a siloed deliverable. You go where the operating problem is.

Autonomous. Embedded is not staff-augmented. You own the method; nobody hands you a ticket.

You won’t start from zero. Provectus builds industry blueprints — working systems that have already shipped for a customer in your industry. Your engagement starts from that baseline, and what you learn in the field goes back into it. That loop is the difference between an outcome and an invoice.

You’ll be measured on whether the Business Unit’s number moved — not on hours, not on scope delivered.

This is a role for engineers who have led before — as a founder, a CTO, a staff engineer — and who want to stay in the code while owning the outcome. On most days you’ll be the most senior technical person in the room, and you’ll still be the one shipping.

Requirements:

8+ years building software, a substantial share of it writing production code you were accountable for. You are hands-on today and intend to stay that way.
You will take the operator’s seat. You are genuinely willing to spend weeks doing someone else’s job — claims processing, underwriting, revenue-cycle work — before you write a line of code. Engineers who need to stay in the IDE should not apply.
You learn domains fast. Demonstrated ability to become conversant in an unfamiliar business function quickly enough to argue with the people who do it for a living.
Shipped GenAI/LLM systems to production — not demos, not notebooks. You’ve handled the parts that get hard after the prototype works.
You evaluate. You have built or owned an eval suite for a non-deterministic system, and you can explain what you measured and why.
Strong engineering fundamentals — dropped into an unfamiliar codebase or language, you’re productive. Python and/or TypeScript proficiency; depth matters more than stack.
Cloud-native delivery on AWS (GCP/Azure a plus): containers, Kubernetes/ECS, IaC, CI/CD, and the operational reality of a system someone else inherits.
Credible with senior stakeholders — you can hold a redesign conversation with a BU head and a scoping conversation with a CTO without losing either room.
Comfort with ambiguity and ownership. Engagements start underspecified by design. Closing that gap is the job.
Solid AI/ML foundations — you understand what the models do well enough to reason about failure modes, not just call the API.
Strong hands-on prodcution experience with Claude Code/Cowork.
Fluent English, written and spoken.
Nice to have:
Prior experience as a founder, CTO, or engineering leader who has chosen to return to individual contribution.
Real depth in one of our blueprint industries: financial services, insurance, healthcare, asset management.
Consulting, professional services, or other embedded customer-facing delivery.
Data platform depth: data lakes, warehouses, streaming and real-time analytics, data mesh and data contracts, governance and data quality.
MLOps and classical ML: PyTorch, SageMaker, MLflow.
Fine-tuning, distillation, or inference/serving optimization.
Graph databases (Neo4j, AWS Neptune).
IaC depth: AWS CDK, CloudFormation, Terraform.
Open-source contributions or public writing on applied AI.

What you’ll do:

Take the seat

Do the operator’s job for two to four weeks at the start of an engagement. Learn the function from inside, not from a requirements doc.

Reach working fluency in a new domain — insurance underwriting, healthcare revenue cycle, asset flow — in weeks, not quarters.

Sit with the operator and the Forward Deployed Executive and redesign the function from first principles. Discovery, user research, and PRD-writing collapse into one team that re-imagines its own job. You are all three roles.

Build

Ship production GenAI systems into the customer’s environment — LLM applications, agentic workflows, retrieval and structured-extraction pipelines, and the services around them. Running software, not recommendations.

Build the evaluation harness before you build the feature. When the engagement is bound to a business KPI, “it looked good in testing” is not an answer. Define what working means, instrument it, and let the evals drive the design.

Write production code across the stack — backend services, data pipelines, and the AI layer. Python and TypeScript are our centre of gravity; we choose tools to fit the customer, not the résumé.

Take systems to production on AWS (GCP/Azure where the customer requires it): containerized, observable, and maintainable after we leave.

Start from the blueprint, and feed the blueprint. What you learn in the field becomes the baseline the next engagement starts from.

Own the outcome

Work in a pair with a Forward Deployed Executive who carries the Business Unit’s KPIs. Your work is measured against the same number.

Drive adoption. A system the BU routes around has not shipped. Change management is part of the engineering job here, not a phase after it.

Be credible with the customer’s engineers, their operators, and their executives — and be willing to disagree with all three.

Shape what we commit to before we commit to it. You’ll have the standing to do it, because you’re the one who will build it.

What We Offer:

Frontier delivery work across Cowork Activation, Agentic SDLC, and Blueprint Activations in Financial Services and Healthcare
The chance to shape how leading enterprises adopt AI, from strategy through first deployment
A forward-deployed model working in small, senior teams alongside Principal Architects and Forward Deployed Engineers
A growing AI delivery practice where you help build the tooling and frameworks, not just use them
Remote-friendly culture

How we hire:

• Intro conversation — the role, your background, what you want to be doing.
• Two live engineering sessions. Real problems, your own editor. You may use an LLM assistant (ChatGPT, Claude) — how you work now includes these tools. Autocomplete/agentic coding tools are off for these sessions.
• The redesign session. We hand you an unfamiliar business function and the constraints of the person who performs it. You have to understand the job well enough to rebuild it — then say what you’d build and how you’d know it worked. No LLMs for this one.
• Team and practice conversation.