Solutions Manager

City Of London, England

Job Summary

Experience

[8 -12 years]

Reports to

[Delivery / Engineering Manager]

Team

Leads a lean pod of 3 agent engineers

About the role

As an Agentic Forward Deployed Engineer, you operate at the front line of delivery - embedded with the client, turning ambiguous business problems into production agents, fast. Your deliverable is Business Transformation Agents: autonomous and multi-agent systems that automate and reimagine real business processes such as invoice disputes, procurement approvals, onboarding, claims and compliance workflows. You own each agent end to end -conceptualize, build, integrate, evaluate, deploy, and sustain - and you lead a small team to do the same. You build exclusively in Python using agent development kits, and you bring Agentic AI capabilities to life inside the client's world, with Responsible AI, evaluation and security as non-negotiables.

Key Responsibilities

What you'll do

Conceptualize fast: embed with stakeholders, frame a business process as an agentic solution, and stand up a working agent prototype in days, not weeks.

Build Business Transformation Agents: design and ship single-agent and multi-agent systems in Python using ADKs that automate and transform real client workflows, with measurable ROI.

Own efficiency as the scorecard: drive delivery efficiency and operational efficiency ; shorter cycle times, less manual effort, higher accuracy, lower cost-to-serve.

Engineer the agent core: apply prompt engineering, context engineering, prompt caching, RAG / context-graph retrieval, memory, tool / function calling, MCP integration and multi-agent orchestration.

Integrate to standards: connect agents into client ecosystems through proven integration patterns, standards-based APIs and secure authentication.

Make reusability and predictability the default: build reusable agent components, skills, tool libraries and templates; add guardrails so agent behaviour is predictable, safe and repeatable.

Prototype and iterate quickly: use the kit's scaffolding to prototype, then harden to production-grade, well-tested Python.

Run eval-driven development: build evaluation harnesses and test suites that measure agent correctness, safety and regression before anything ships.

Own AgentOps / DevSecOps: CI/CD for agents, versioning, observability and telemetry, shift-left security, and Responsible AI governance baked in from day one.

Run a continuous, adaptable feedback loop: feed production telemetry, evals and client feedback back into prompts, context and agent design.

Stay ahead of the curve: adopt evolving agent frameworks and patterns quickly, and bring field learnings back to the practice.

Lead and mentor: set technical direction for a lean team of 3 agent engineers, raise the engineering bar, and grow the pod's agentic capability.

What you'll bring (must-have)

Strong Python engineering ; idiomatic, typed, tested and packaged code; on a foundation of solid software engineering principles (design, version control, architecture).

Hands-on agent building with at least one agent development kit (Google ADK, LangGraph, CrewAI, OpenAI Agents SDK, AWS Bedrock AgentCore or Microsoft Agent Framework / Semantic Kernel).

Solid command of agent engineering: prompt engineering, context engineering, prompt caching, RAG / context graphs, tool / function calling, MCP, and multi-agent orchestration.

Eval-driven development: designing evaluation harnesses and measuring agent quality, safety and reliability.

Standards-based integration and DevSecOps: APIs, secure auth, CI/CD, observability and AgentOps.

Ability to conceptualize a business problem as an agent quickly, and operate effectively in ambiguous, customer-embedded settings.

Client-facing maturity: translates fluidly between technical and non-technical stakeholders, and owns outcomes.

Experience mentoring or leading small engineering teams.

Skill Requirements

Technology mandate

Language: Python preferable

Frameworks: Agent Development Kits (ADKs) ; e.g. Google ADK, LangGraph, CrewAI, OpenAI Agents SDK, AWS Bedrock AgentCore, Microsoft Agent Framework / Semantic Kernel. Framework choice follows the engagement; the discipline is the same.

Models: Multi-LLM via the kit (e.g. Claude on Bedrock, Gemini, Azure OpenAI), selected per use case for quality, latency and cost.

Interfaces: Tools and Model Context Protocol (MCP) for integration; standards-based APIs and secure auth for client systems.

What great looks like (strongly preferred)

Fluency across multiple ADKs and the judgment to pick the right one per engagement.

Deploying agents to managed runtimes at enterprise scale (e.g. Vertex AI Agent Engine, Bedrock AgentCore) with governance and cost control.

Domain depth in a transformation area - finance operations, supply chain, HR, claims or compliance.

Experience with an enterprise agent platform, including Responsible AI and governance at scale.

A track record of turning agents into reusable accelerators or IP adopted beyond a single engagement.

#body.unify div.unify-button-container .unify-apply-now: focus, #body.unify div.unify-button-container .unify-apply-#body.unify div.unify-button-container .unify-apply-now: focus, #body.unify div.unify-button-container .unify-apply-