Back to all positions

Forward Deployed Engineer

(AI Solutions Engineer, Delivery Lead)

Level: Senior / Staff8+ years engineering, 3+ shipping production systemsRaleigh, NC. This is a Raleigh based role.Reports to Brad Springer, CEO, Bob.pro

You join as an early member of the engineering team.

The 30 second version

We build governed AI operators, which we call BOBs, that do real work inside small and mid market businesses. Not copilots. Not chat wrappers. Systems with role contracts, budget limits, audit trails, and a human who can see exactly what the system did and why.

You will own client deployments end to end, from the first discovery conversation through production and measurement, working alongside a senior CX and Product partner who owns the client relationship. What you learn in each deployment goes back into the platform so the next one starts further ahead.

Why this role exists

We are a small team with production clients and a growing pipeline. We are good at the engineering and we are good at the outcomes. What we need now is another person who can hold the whole arc: hear a business problem directly from the operator who has it, design the system that solves it, write the code, put it in production, prove it worked, and then make the platform smarter so the next deployment goes faster than the last.

Most engineers can do the middle of that arc. Fewer can do the front of it. Almost nobody does the back of it unless asked. We are hiring for all three.

You will also help guide the technical bar for the rest of the engineering team.

What you will actually do

Roughly how a month breaks down, so you can decide whether this is the job you want:

~50%

In a repo

You write the code. This is not an architecture role with a hand off to someone else. If you have not shipped production code in the last eighteen months, this is the wrong role.

~25%

Client facing

Discovery sessions, workflow mapping, design reviews, go live, and the conversation where you show an owner what the system did last week and what it cost. Typically two to three concurrent accounts.

~15%

On the platform

Turning what you just learned into platform capability: identity and state, tenant scoped memory, playbook construction, tool and action surfaces. Configuration over code forks, every time.

~10%

On leadership

Everyone here is expected to lead from the front: design review for other engineers, holding the delivery standard, mentoring, and setting the example by doing the work well. Your influence comes from what you ship and how you handle yourself in front of a client, not from a reporting line.

How you work with CX and Product

You have a partner. A seasoned CX and Product leader who owns the client relationship: agenda, cadence, expectations, escalation, follow up, solution design and commercial temperature. That person runs the room.

You are in the room. Not on the recording, not reading the readout. For discovery and design review you are present and asking your own questions, because a requirement that reaches you secondhand has already lost the detail that mattered.

Yours

Technical discovery, the code, deployment, evals, guardrails, instrumentation, and the technical relationship with the client's operator and their technical counterpart.

Not yours

Renewals, pricing, upsell, invoicing, change order negotiation, quarterly business reviews, and the tier one support queue. You flag commercial risk. You do not carry the quota or the relationship maintenance.

We draw this line on purpose. We have seen what happens to engineering throughput when we do not.

The technical work

You will design against four primitives.

Identity & State

Every interaction anchored to objects under strict org level isolation.

Memory & Context

Tenant scoped transcripts, documents, embeddings, and artifacts in partitioned data planes.

Reasoning & Playbooks

Model instructions constructed dynamically from each tenant's methodology, ICP, and vocabulary. No fixed framework unless the client asks for one.

Actions & Surfaces

Tenant specific tools loaded through an orchestrator. Configuration over code forks.

Our production bar, which is not negotiable:

Test coverage that means something: unit, integration, end to end
Versioned schema migrations. Nobody touches a production database by hand.
Tenant isolation enforced at every layer, verified rather than assumed
Secure by default APIs and infrastructure
Documentation updated as part of definition of done
Every agent action attributable, budgeted, and reversible

We optimize for velocity under constraint. Not speed without discipline, and not discipline without shipping.

What we are looking for

Required:

8+ years building software, including 3+ years shipping and operating production systems you were personally accountable for
You have shipped LLM powered systems that real users depend on: RAG, structured agents, tool calling workflows, whatever the shape
Multi tenant architecture experience. Isolation, data partitioning, per tenant configuration, and the failure modes of all three.
You have built evaluation into an AI system. Offline evals, shadow mode, scoring, regression suites. Not “it looked good in testing.”
You have built guardrails and rollback paths, and you have used them under pressure
You can run a technical discovery conversation with a non technical business owner and leave with a design. This is a hard requirement and we test it directly.
You write clearly. Short, structured, unambiguous. We read your writing before we read your code.
You have tied a system you built to a number a business cared about, and you can tell us what happened when it moved, or when it did not

Nice to have. Genuinely nice to have, not a hidden filter:

Hands on with OpenClaw, the Claude Agent SDK, MCP, or comparable agent runtimesForward deployed, consulting, or small shop background where you were the technical half of the engagementSMB or mid market domain depth: field services, healthcare, professional servicesExperience turning delivery patterns into platform capability

Who this is wrong for:

Researchers. We are not doing novel work. We are doing correct work, quickly, for people whose payroll depends on it.
Architects who want to specify and delegate.
Anyone who needs a written spec before they start.
Anyone who wants to work only with technical stakeholders. Our clients run lawn care companies, medical practices, and nonprofits. You will explain retrieval boundaries to someone standing next to a truck.

What you get

Health coverage through Sedera, paired with Direct Primary Care.

Membership based care with a doctor you can actually get on the phone, not a carrier phone tree.

Unlimited PTO.

Take the time. We care that the work ships and the clients are covered.

Real authority over how we deliver, on day one, in writing.

You own the delivery engineering standard, you decide how deployments run, and nobody above you is going to relitigate your architecture in a hallway.

Surface area you will not get at a larger company.

You will touch every layer, talk to every client, and see the direct line from what you built to what the business did.

How to apply

Send to careers@bob.pro:

Your resume
One page about a production AI system you shipped. What it did, who used it, how you knew it was working, and what broke. We care more about the last paragraph than the rest combined.

If that page is interesting, the process is:

Step 1

30 minutes with Brad

The role, the company, compensation, your questions.

Step 2

Meet the team

You sit with the engineers you would be working next to and with your CX and Product partner. They ask their own questions and their read carries real weight. We are small enough that everyone here works with everyone.

Step 3

90 minute work session

We hand you a real, anonymized client problem and you work it. When you are done we look at the repo together to see what you built. We are watching how you worked the problem, not whether you finished it.

Step 4

Reference calls and offer

Four steps, 2 weeks, and you will know where you stand after every one.

careers@bob.pro

Your privacy choices

Essential storage remembers this choice. With your permission, analytics also preserves bounded campaign attribution and loads Google Tag Manager and Apollo. Booking links still work if you decline. Read the privacy policy.