StarApple AI · Applied engineering programme

AI's future is built in the field.

Forward Deployed Engineers bring AI into the real world: working with people, understanding their workflows, and building systems that solve practical problems.

Explore the programme

Proposed eight-week programme · Project-led · Dates and fees to be confirmed

First, what is an FDE?

A Forward Deployed Engineer works alongside users to turn a real operational need into software that can be tested, adopted and improved.

FDEs combine software engineering with discovery and implementation. They go where the problem lives: learning the workflow, connecting AI to the right systems, measuring whether it helps, and preparing the people who will run it.

That is the value: less distance between an AI idea and a useful, supportable product. The role is emerging as organisations move beyond experiments and need builders who can deliver in context.

Read our field guide to the FDE role →
The FDE field loop

Stay close to the problem. Own the outcome.

The job spans the full path from first conversation to a system another team can confidently operate.

  1. 01 / DISCOVERUnderstand

    Listen to users. Map the work and the friction.

  2. 02 / DESIGNChoose

    Match the approach to data, risk and constraints.

  3. 03 / BUILDIntegrate

    Make the AI feature work within the real workflow.

  4. 04 / PROVEEvaluate

    Test quality, safety and usefulness with evidence.

  5. 05 / TRANSFEREnable

    Prepare monitoring, ownership and next steps.

What you will be able to do

Build beyond the demo.

Practise the technical and human work that makes AI useful in a real organisation.

Find the right problem

Map a workflow with users, identify constraints and define a measurable baseline.

Engineer the system

Connect models, data, retrieval and tools to the applications people already use.

Make it dependable

Evaluate results, protect sensitive data, design human oversight and plan the handover.

Proposed eight-week curriculum

One field problem, built in stages.

Open a week to see its focus and practical artefact. Every stage feeds the capstone.

WEEK 01Discover the workflow

Interview users, map the current process and identify where AI may help. Set a baseline and agree what success means.

  • User and workflow map
  • Failure points and data needs
  • Baseline and acceptance test

Field artefact: one-page problem brief.

WEEK 02Design data and architecture

Inspect data quality, permissions and formats. Compare rules, machine learning, retrieval and generative AI against the actual need.

  • Data and API inventory
  • System architecture diagram
  • Model, cost and dependency decision

Field artefact: architecture and data plan.

WEEK 03Build retrieval and context

Prepare documents and build a retrieval workflow. Test citations, missing information and responses that should be declined or escalated.

  • Document ingestion and retrieval
  • Structured output and attribution
  • Tests for incomplete or conflicting records

Field artefact: grounded assistant using practice data.

WEEK 04Integrate with an application

Build a usable interface and service boundary. Work with APIs, validation, authentication, state, retries and clear error handling.

  • Input and output contracts
  • Application integration and access controls
  • Fallbacks for service failures

Field artefact: working workflow slice.

WEEK 05Orchestrate tools responsibly

Build tool-using workflows and learn when a predictable sequence is better than an agent. Set permissions, approvals and action limits.

  • Typed tools and input validation
  • Human approval for consequential actions
  • Failure, retry and escalation paths

Field artefact: supervised tool workflow with traces.

WEEK 06Evaluate quality and failure

Build a representative test set. Score task success, groundedness, format, latency and cost. Add regression and adversarial tests.

  • Evaluation rubric and labelled set
  • Quality, safety and cost results
  • Failure analysis and fixes

Field artefact: evaluation report with evidence.

WEEK 07Prepare a controlled pilot

Review privacy, security, access, logging and ownership. Add monitoring and user feedback. Plan how to pause or recover from failure.

  • Threat and risk register
  • Release checklist and monitoring plan
  • Cost and value measures

Field artefact: pilot readiness pack.

WEEK 08Present and hand over

Demonstrate the workflow, tests and limits. Gather feedback, explain operations and recommend a clear next decision.

  • Working pilot or testable prototype
  • Evaluation evidence and risk controls
  • Runbook, owner map and next-step plan

Field artefact: capstone demonstration and handover.

Assessment and completion

Evidence matters as much as the build.

The capstone is assessed on whether the work fits the problem, performs reliably and can be responsibly operated.

Capstone project

Choose a bounded workflow challenge and build a pilot or testable prototype using synthetic data or data approved for the learning environment. Example practice scenarios include document search, service intake, claims triage or internal knowledge support. These are learning exercises, not StarApple AI client results.

Present the problem, architecture, working feature, evaluation, limits, risk controls and handover plan.

How it is assessed

  • Problem fit and user evidence20%
  • Engineering and integration20%
  • Evaluation and test evidence20%
  • Safety, privacy and security20%
  • Operations, value and handover20%

Completion requires at least 70% overall, 80% attendance and a pass on the safety and privacy gate. Successful learners receive a StarApple AI Certificate of Completion. This is not an external licence or accredited degree.

Is this for you?

For builders ready to work close to users.

A good fit if you…

Build software, work with data, shape technical products or bring deep knowledge of a domain.

  • Have basic Python or JavaScript experience
  • Know the basics of HTTP, JSON, APIs and Git
  • Want to test ideas with users and explain trade-offs

New to AI engineering?

Prior production AI experience is not required. The programme starts with problem framing and builds toward more complex engineering. If you are new to coding, tell us so we can recommend preparation before enrolment.

  • Bring a laptop and a problem you can safely explore
  • Tools are confirmed for each cohort and organisation
  • Never use confidential data in unapproved AI tools
Technical references and guidance

The curriculum is informed by these public frameworks; they do not certify the programme.

AWS Generative AI Lifecycle Operational Excellence informs the progression from experimentation to validation and production operations.

OpenAI guide to building agents and agent evaluation guidance inform tool workflows, tracing and evaluation.

NIST AI Risk Management Framework and its Generative AI Profile inform risk identification and measurement.

OWASP GenAI Security Project: Top 10 for LLM Applications (2026) informs application security and abuse testing.