Principal Software Engineer

Work at the intersection of enterprise data, business semantics, AI, workflows, security, and human decision-making. Shape a platform while its most important contracts and operating model are still being defined. Build technology that becomes part of how enterprises operate, rather than isolated demonstrations or thin model wrappers.

Responsibilities:

Set and evolve the reference architecture across the Lighthouse control plane, Context Layer, data and execution adapters, workflow and agent runtime, evaluation, and production operations.

Define the contracts that keep the platform portable: canonical business concepts and tenant mappings; reusable Context Modules and task Recipes; context requests, plans, packages, and traces; and the runtime boundary between context compilation and data execution.

Own security and deployment architecture for shared, dedicated, and customer-hosted environments, including identity, authorisation, tenant isolation, secrets, network boundaries, data residency, audit, and policy enforcement before retrieval and action.

Design reliable execution for deterministic and model-driven workflows, including long-running state, approvals, retries, idempotency, timeouts, partial failure, compensation, safe replay, and rollback.

Establish engineering standards for AI quality and system operations: evaluation datasets, regression gates, tracing, evidence and grounding, service-level objectives, incident learning, latency and cost budgets, and model or tool fallback.

Make build-versus-buy and open-source extension decisions with clear upgrade paths; turn proven patterns into versioned services, APIs, connectors, SDKs, and solution templates rather than premature abstractions or customer-specific forks.

Personally design, implement, and review the components with the greatest architectural, security, performance, or operational risk; mentor engineers and communicate decisions clearly to product, business, and customer stakeholders.

Expectations:

Lighthouse has clear system boundaries and stable contracts, so new use cases reuse platform capabilities without creating hidden coupling or one-off forks.

Context is permission-aware, bounded, testable, and explainable: the system can show which business definitions, sources, policies, freshness checks, and evidence supported an answer or action.

Production workflows degrade safely and recover predictably, with measurable reliability, quality, latency, cost, and business outcomes.

The engineering team moves faster because high-impact decisions are documented, risks are surfaced early, ownership is clear, and standards are enforced through code and tooling.

Requirements:

8+ years building production software, including principal-level technical leadership across distributed, data-intensive, platform, or enterprise systems.

Deep hands-on strength in Python, API and service design, relational data modelling, asynchronous and event-driven systems, and cloud-native operations.

Experience defining architecture across multiple services and deployment modes, including identity, authorization, multi-tenancy or customer isolation, observability, and failure recovery.

Experience taking AI or ML systems beyond prototypes, including structured model outputs, tool use, retrieval or context, evaluation, guardrails, and production debugging.

Strong judgment about system boundaries, data ownership, consistency, scalability, security, build-versus-buy choices, and how much abstraction is justified at each stage.

A track record of leading through influence while staying close to implementation, code quality, production behaviour, delivery sequencing, and customer outcomes.

Useful, but not required: Experience with enterprise metadata, semantic or knowledge-graph systems, workflow engines, data platforms, developer platforms, or multi-tenant SaaS.

Familiarity with parts of our likely stack: FastAPI/Pydantic, PostgreSQL, Redis, queues or event systems, MCP, vector or graph stores, Docker/Kubernetes, and AWS, Azure, or GCP.

Experience integrating or extending open-source platforms while preserving upgrade paths, or deploying into regulated, private-network, or customer-controlled environments.