Forward Deployed Engineer – Professional (CFDE-P)

Strategic sponsorship and initial stakeholder alignment

Pre-project strategy. Business case development, roadmap creation, resource allocation. Securing early sponsorship for the AI initiative. Planning answers “how do we get the green light.” Distinct from Module 13 (Adoption) which covers “how do we sustain the change.”

Legacy databases, ERPs (SAP, Oracle), CRMs, mainframes, undocumented APIs

Connecting AI systems to existing enterprise infrastructure. Working within constraints you did not design and cannot change. Data migration, ETL pipelines, format transformation. The hands-on “how to connect AI to what already exists” module.

SSO, SAML, OIDC implementation. Zero-trust architectures.

The practical reality of getting production credentials from an enterprise security team. Identity federation across systems. This is the skill that separates enterprise FDEs from startup FDEs.

Lab session from Sovereign AI Lab (SAIL)

Deploying AI systems inside client-controlled infrastructure with restricted or no internet access. Government, defence, and financial services environments. Private cloud, on-premise GPU clusters, air-gapped networks. Packaging models and dependencies for offline deployment. Testing without external API calls. SAIL lab provides a hands-on component that no other certification offers.

Adversarial prompt testing, data poisoning detection, prompt injection defence. AI supply chain security.

Technical compliance: applying controls, hardening. AI supply chain treated as a security function: model and dataset provenance tracking, secure model registries, third-party model risk assessment, open-source model vetting. Distinct from Module 10 (Governance) which covers operational governance and regulatory navigation.

Incident response. Model risk, reputational risk, vendor risk. Bias and fairness. Multi-agent emergent behaviour. Safety engineering.

Understanding what can go wrong with AI systems and how to prevent, detect, and respond to it. Bias and fairness covers the audit methodology, regulatory expectations, and what biases to look for. Technical testing execution lives in Module 11 (QA).

Token and GPU cost modelling. Chargeback and showback. Business case for AI investment.

The financial discipline required to keep enterprise AI solvent. Per-token billing, GPU scarcity, volatile pricing. Cost baselining, GPU optimisation, quota enforcement, cross-team chargeback. Without this, enterprises suffer “agent sprawl” and surprise million-pound inference bills that freeze further AI investment.

Output drift detection. SLA compliance monitoring.

Logging, alerting, and monitoring for systems serving thousands or millions of users. Token usage tracking and cost attribution across business units. Latency monitoring and SLA compliance. Tools: LangSmith, Langfuse, Arize Phoenix, Datadog. Building observability that the client’s operations team can own after you leave.

Agent versioning, deprecation, retirement. Permission scoping. Staged rollouts.

Not building agents (that is Foundation Module 10). Managing the full lifecycle: versioning, permission scoping, staged rollouts from supervised to semi-autonomous to fully autonomous, deprecation, retirement. Agentic workflows need their own lifecycle beyond traditional MLOps.

Governance-as-code, kill switches. EU AI Act, NIST AI RMF, UK AI Safety Institute. Audit trails. Incident reporting.

Building governance structures for the client: who approves model changes, who audits, who has override authority. Documentation requirements for regulated deployments. Preparing for regulatory inspections. Distinct from Module 5 (Security/Compliance) which covers technical controls. This is operational governance and regulatory navigation.

Model validation, quality gates, fairness testing, client acceptance criteria

Systematic quality engineering between observability (which monitors) and governance (which enforces policy). Test strategy, validation datasets, performance thresholds, QA processes across the AI lifecycle. Fairness testing here is the technical execution: tooling, metric selection, integration into CI/CD.

Timelines, milestones, dependency management. Sprint cadence. Status reporting. Commercial awareness.

Execution mechanics. Different from Module 13 (Adoption/Change Management). Project Management answers “how do we deliver this on time and on budget.” Commercial awareness ensures the Professional cert graduate understands that AI projects are business instruments, not just technical deliveries.

Coalition building, AI literacy programme design (EU AI Act mandate), overcoming workforce resistance, adoption metrics

The human operating system. Ground-level resistance, fear, coalition building, AI literacy, cultural shift. EU AI Act mandates AI literacy as a legal compliance requirement, making this a regulatory obligation. Distinct from Module 1 (Planning) which covers pre-project strategic sponsorship. This is ongoing transformation during and post-deployment.

Requirements elicitation from resistant stakeholders. Specification and validation.

How to elicit, document, and translate messy business needs into technical specifications. Technical at its core while acknowledging the human dimension: resistant stakeholders, conflicting priorities, unspoken constraints. An established discipline (IEEE, IREB certify it) applied specifically to AI deployments.

Product feedback loops: field patterns to platform features. Repeatable deployment patterns and internal playbooks.

The mechanism by which the enterprise learns from live AI and hardens it. Identifying repeatable deployment patterns. Codifying field insights into internal documentation. Contributing to product roadmap from the field. The skill that makes an FDE more valuable than a consultant: you improve the product, not just the deployment.

Executive communication and business impact

Full enterprise engagement simulation. Realistic brief with legacy system constraints (mock SAP/Oracle environment), compliance requirements (SOC 2 evidence, data residency), multiple stakeholders with conflicting priorities (CTO wants speed, CISO wants security, CFO wants cost reduction), ambiguous requirements, and a tight deadline. Deliverables: discovery documentation, architecture design, working prototype integrated with mock enterprise systems, compliance documentation, deployment plan, executive presentation, and product feedback report. Assessed by practitioner panel via 45-minute oral defence. Graded on: technical quality, enterprise-readiness, stakeholder management, communication under pressure, and commercial awareness.