Senior Software Engineer

Pearson•United States,

16d•$90,000 - $135,000

About The Position

Pearson, the world’s leading learning company, is hiring a Senior Software Engineer to play a key role on our AgentOps Engineering team. This is a hands-on senior engineering role focused on building. You’ll ship the write-path features that let agents take real actions, design the orchestration and routing patterns that direct work to the right agent, and engineer enhancements to the platform itself—working inside it, not just on top of it. Through this, you’ll help teams across Pearson put dependable AI agents to work, and help people realize the lives they imagine through learning. As a key member of the AgentOps Engineering team and Pearson, you will be expected to embody, and role model our leadership dimensions: Live Our Purpose, Simplify the Complexity, Carry Our Culture, and Deliver Results.

Requirements

A hands-on senior engineer who builds production-grade agentic systems, not demos.
An independent owner who takes complex workstreams from ambiguity to reliable, shipped software.
A collaborator who influences through standards and mentoring, without needing authority.
Someone with strong taste and trade-off judgment across quality, latency, cost, resilience, and maintainability.
You build LLM-powered systems, agents, or digital workers that run in production.
Strong Python and backend/platform engineering—async services, typed code, and clean architecture.
You work fluently with agent orchestration frameworks and can design, build, and orchestrate reliable agent tools—contracts, error handling, tool chaining.
You’ve built routing, dispatch, or workflow-control logic that directs work across components or services.
You use AI-pair-programming as a primary delivery mode without sacrificing quality.
Strong prompt engineering for structured outputs, nested schemas, and multi-agent coordination.
You read, edit, and contribute to a real platform codebase—APIs, runtime, storage—not just an SDK.
Solid APIs, distributed systems, and cloud-native engineering, with production-reliability instincts.

Nice To Haves

Experience evaluating agents—task success, groundedness, tool-use accuracy, schema conformance, and regression against golden fixtures.
Experience with tool/context interoperability protocols (such as MCP).
Experience with a major cloud platform (AWS preferred), containerization, and CI/CD; familiarity with state stores for orchestration and persistence.
AI observability and evaluation tooling for LLM systems.
RAG and memory patterns: vector databases, hybrid retrieval, re-ranking, and grounding.
Secure execution, sandboxing, and prompt-injection mitigation.

Responsibilities

Build write-path features: Ship the features that let agents take real, governed actions—wrapping enterprise APIs, validators, and data sources so agents can write, not just read. Design safe write semantics: idempotency, confirmation and human-in-the-loop gates, structured outputs, and audit-ready action logging.
Design multi-agent orchestration: Build multi-agent and digital-worker orchestration patterns that let specialized agents delegate, collaborate, and complete multi-step goals. Build stateful, cyclic workflows using modern agent orchestration frameworks—enabling reflection, recovery, and adaptive execution beyond linear chains.
Engineer agent routing: Build the routing layer that directs requests to the right agent or crew—intent classification, capability-based dispatch, fallback, and escalation paths. Create reusable components for retries, degraded modes, and human handoff; tune agent roles, goals, and prompts using fixtures and golden sets.
Enhance the platform: Work hands-on with the agentic platform—creating, versioning, invoking, and debugging crews, tasks, and graphs through platform APIs—and contribute enhancements back where new patterns are needed. Design long-running, resumable workflows (checkpointing, persistence, context restoration) and resilience patterns for non-deterministic AI.
Ship with AI, at quality: Work primarily through AI-pair-programming as your default mode—researching, scaffolding, implementing, reviewing, and shipping at high velocity while holding quality through TDD and rigorous verification.
Raise the bar: Drive reusable engineering standards, shared libraries, and reference patterns; mentor engineers through design and code reviews; and share the prompts, skills, and workflows that multiply everyone’s output.