Batch 1 (Nov 2026 Intake) — Starts Thu, 12 Nov 2026

What you'll learn

1. 1. Weeks 1–2: FDE Discovery & Rapid Enterprise ScopingThe Forward Deployed AI Engineer (FDE) operating model (Palantir / OpenAI / Anthropic field engineering playbook) · Translating ambiguous C-suite & business pain points into quantifiable AI architecture specs within 48 hours · Defining production SLAs: p95 latency budgets, token cost ceilings, security boundaries, and deterministic fallback paths · Enterprise data readiness audits: evaluating messy schemas, OCR quality, access controls, and build-vs-buy trade-offs
2. 2. Weeks 3–4: Messy Enterprise Data Ingestion & Hybrid RAG in the FieldIngesting real-world enterprise mess: multi-column PDFs, scanned invoices, SQL/ERP exports, SharePoint & Confluence wikis · Document-level Role-Based Access Control (RBAC) & ACL metadata filtering so users only retrieve what they are authorized to see · Hybrid retrieval architecture: Dense embeddings + BM25 lexical search + Reciprocal Rank Fusion (RRF) + Cross-Encoder Reranking · Strict inline citation enforcement, table-aware chunking, and 'I don't have enough verified context' abstention logic
3. 3. Weeks 5–6: Agentic Workflow Automation & Custom MCP/Tool IntegrationArchitecting stateful multi-agent workflows with LangGraph and Google ADK (Supervisor, Router, and Human-in-the-Loop patterns) · Building custom Model Context Protocol (MCP) servers and typed tool-calling interfaces for enterprise systems · Integrating agents with live business systems: Jira, ServiceNow, Salesforce/HubSpot CRM, Slack, and internal REST/GraphQL APIs · Idempotency, transactional rollbacks, cyclic error recovery, and approval gates for high-stakes write actions
4. 4. Weeks 7–8: LLM Evals, Guardrails, Red-Teaming & Zero-Hallucination SLAsDesigning golden evaluation datasets from real client failure logs and edge-case transcripts · Automated CI/CD evaluation pipelines: Faithfulness, Context Recall, Tool-Selection Accuracy & LLM-as-a-Judge rubrics · Adversarial red-teaming: defending against indirect prompt injection, jailbreaks, data exfiltration, and toxic outputs · Enterprise compliance middleware: real-time PII redaction, audit logging, and Google Cloud SAIF security alignment
5. 5. Weeks 9–10: VPC/On-Prem Deployment, CI/CD, Observability & Latency OptimizationContainerizing production AI microservices with Docker, multi-stage builds, async workers, and health checks · Deploying across enterprise environments: Google Cloud Run, Vertex AI endpoints, and private VPC / vLLM self-hosted patterns · Full-stack LLM observability: OpenTelemetry & Langfuse distributed tracing, token cost attribution, and drift alerts · Cutting latency & cloud bills by 60%: semantic caching, prompt compression, model routing (Flash vs Pro), and streaming SSE
6. 6. Weeks 11–12: Live FDE Client Simulation Capstone & Executive Sign-OffEnd-to-end 2-week simulated enterprise deployment from raw client brief to production cutover · Week 11: Client discovery call, data audit, Architecture RFC approval, and full ingestion + agent + eval implementation · Week 12: Production deployment, executive ROI readout deck, runbook handover, and live stakeholder Q&A defense · FDE Career Playbook: Cracking Forward Deployed Engineer interviews (coding, messy-data debugging, and live client roleplay rounds)

Final project demo: End-to-End Enterprise Forward Deployed AI System & Executive Sign-Off. Act as the Lead Forward Deployed AI Engineer for a simulated enterprise client (e.g., BFSI Claims Triage & Policy Copilot, Healthcare SOP & Compliance Agent, or Manufacturing Supply-Chain Incident Resolver). Deliver the Architecture RFC, RBAC Hybrid RAG pipeline, custom MCP tool-calling agent, automated evaluation suite, Dockerized Cloud Run deployment, and live executive sign-off presentation.