FORWARD DEPLOYED AI ENGINEERING · ENTERPRISE AI
Jorge Carbonell
I embed with your team and customers to take artificial intelligence from pilot to production. I build AI agents, RAG and LLM applications around real data and workflows; backend, integrations and cloud provide the foundation to operate them.
01 / WHERE THE BUSINESS GETS STUCK
A model is not enough: enterprise AI needs accessible data, workflow integration, evaluations and operations. As an FDE, I work with your team to address these blockers.
01 /
Test data looks great. Real permissions, failures and users expose the gaps.
WHAT I IMPLEMENT
A scoped end-to-end path: real data, authentication, evaluations, observability and rollback.
WHAT WE TRACK
Testable acceptance criteria and a launch-readiness checklist.
02 /
SSO, APIs, ERP and data requirements delay activation and consume engineering capacity.
WHAT I IMPLEMENT
Reusable connectors, API contracts and deployment patterns, alongside the customer's technical team.
WHAT WE TRACK
Time to the first operational workflow and open integration blockers.
03 /
Hallucinations, out-of-scope data and overprivileged agents hold back real usage.
WHAT I IMPLEMENT
Access-aware RAG, real-world evaluations, traces, tool boundaries and human review where it matters.
WHAT WE TRACK
Evaluation-set quality, failures and human escalations.
04 /
Product, IT, security and business work separately. Launch falls between teams.
WHAT I IMPLEMENT
A technical point of contact who builds, documents decisions and makes dependencies and exit criteria visible.
WHAT WE TRACK
Open dependencies, owners and decisions needed to launch.
05 /
Bespoke solutions accumulate. Every deployment adds maintenance that is hard to sustain.
WHAT I IMPLEMENT
Separate configuration from custom code, extract shared components and feed field signals back to product.
WHAT WE TRACK
Reused components, documented exceptions and maintenance burden.
06 /
A new tool won't change a process if it forces people to work differently.
WHAT I IMPLEMENT
Embed in existing workflows, validate with users and prepare operations and team handover.
WHAT WE TRACK
Repeat workflow usage, completed tasks and operator feedback.
02 / THE MISSING ROLE BETWEEN TEAMS
An FDE combines customer context with hands-on execution. I don't replace your product, security or sales teams: I work with them and bring the solution into the field.
Discovery with real users, systems and constraints. Not a brief passed between teams.
Code, APIs, data and deployments. Decisions visible in repositories, not just slide decks.
Feedback, reusable patterns and documented boundaries so the next customer doesn't start from scratch.
Connect internal documents and data to LLMs using retrieval-augmented generation (RAG). Build context retrieval, citations and access permissions; evaluate relevance and answers against real queries.
Build agents that use APIs and tools across support, operations and internal workflows. Define permissions, execution boundaries, retries and human approval for sensitive actions.
Integrate language models into existing products and tools. Build structured outputs, model selection and per-task quality, cost and latency evaluation.
Prepare regression tests, traces and output monitoring. Build GCP, AWS or Azure deployments with scalable backends, alerts and recovery procedures.
03 / WHERE THIS ROLE FITS
Illustrative engagement scenarios, not client case studies or promised results.
For product teams and CTOs who want to activate customers without turning every onboarding into a roadmap detour.
HOW WE TRACK PROGRESS
Time to the first operational workflow, synchronization failures and remaining manual work.
04 / DELIVERABLES, NOT BLACK BOXES
Reviewable changes, tests and documented architectural decisions.
Deployment, alerts, metrics and recovery procedures.
Context handover, known boundaries and prioritized next steps.
We define baselines and targets together. These are progress indicators, not guaranteed improvements or invented ROI percentages.
First we scope a need. Then we build a useful end-to-end path and validate with its operators. Scope and pace depend on access, dependencies and complexity.
Work with the customer and team to map the problem, dependencies and success criteria.
Build an initial solution using real data and workflows; iterate with the people who will use it.
Document decisions, deployment and operations; feed learning back to product and engineering.
15+ years in engineering and 4+ apps in production. My focus: backend, microservices, integrations, GCP, AWS, Azure and applied AI. I work inside your systems and alongside the people using them.
Need to unblock an initiative or bring this role into your team? We can discuss either.
PROJECTS
To implement an integration, deploy AI or take a customer solution to production. Scope, commitment and terms are agreed after reviewing the context.
Discuss a projectEMPLOYMENT
For companies hiring an in-house Forward Deployed Engineer, close to customers and engineering. Let's discuss responsibilities, team, location and salary range.
Offer a roleResponsibilities may overlap between companies. In my offering, beyond architecture I write code and integrations and prepare deployment and operations alongside the customer. The responsibility doesn't end with a design or demo.
They work alongside customers to turn a use case into an integrated, operable AI application. My focus is RAG, agents and LLMs: connecting data and tools, evaluating outputs and preparing deployments with traceability and human oversight. This fits AI-first companies and enterprises adopting AI in existing workflows.
I embed with product and engineering and collaborate with customer IT, security and users. We scope my technical responsibility without replacing business, security or legal-review owners.
A concrete problem, a technical point of contact and agreed access to the necessary people and systems. Start by emailing customer context, the bottleneck and your stack; don't send credentials or sensitive data.
Yes. For projects we agree scope, commitment and terms. For employment we discuss responsibilities, team, working arrangement, location and salary range. No predefined rate or availability.
We can agree an NDA before reviewing sensitive information. We use bounded access and document permissions and decisions. Security and legal reviews belong to your organization's owners; I don't issue certifications.
JORGE CARBONELL / FORWARD DEPLOYED ENGINEER
Tell me the bottleneck, your stack and who's affected. I'll reply personally to assess the fit; no long form or generic proposal.
jorge@weddedgroup.com