Senior Software Engineer

Thomson Reuters•Frisco, TX

1d•$102,200 - $189,800•Hybrid

About The Position

The CoCounsel Legal Integrations team within Thomson Reuters is seeking Senior Software Engineers to join the Content Tools team in building purpose-built APIs that give AI agents precise, governed access to TR’s proprietary legal content across various international jurisdictions. These tools are the bridge between TR’s authoritative content and the AI products, such as CoCounsel Legal, that depend on them, and they carry high stakes — every result must be accurate and grounded in TR’s authoritative content. This is a full-ownership engineering role on a team that designs, develops, tests, deploys, and operates the production content APIs that agentic workflows rely on. The team’s operating principles are non-negotiable: you own what you build end-to-end, from the first commit to the production dashboard. You ship to production constantly, and you treat delivery friction as an engineering problem to solve, not a fact of life. You use AI-assisted development as a primary tool, with the expectation that the majority of code is written with AI assistance. And you practice verified spec-driven development: you write the acceptance criteria and utilize evaluation datasets that specify a tool’s behavior before you build it, and that eval suite gates every change in CI. Senior Engineers own a tool, subsystem, or content domain end-to-end, hold design-decision authority within their scope, serve as the team’s subject-matter expert in at least one area, and mentor more junior colleagues. This is a role for someone who takes genuine pride in the quality, reliability, and pace of their work.

Requirements

Bachelor’s or Master’s degree in Computer Science, Engineering, or a related field
5+ years of software engineering experience designing, developing, and testing production backend, API, or data-infrastructure systems, with demonstrated end-to-end ownership of features and components
Mastery of Python and strong fundamentals, including API design and a rigorous testing discipline (unit, integration, and evaluation) — you write high-quality production code and test everything you ship
Solid distributed-systems knowledge: you understand the failure modes of services, queues, and datastores and design systems resilient to them
Proficiency with AWS cloud services , Datadog and CI/CD practices (Blue/Green, Canary deployments); you ship small PRs frequently with fast feedback loops
Demonstrated operational ownership: on-call participation, architectural documentation and decision records, post-mortems, and a clear philosophy that building and operating are inseparable
Fluency with AI-assisted development tools as a primary part of your workflow — you write the majority of your code with AI assistance and use it to accelerate delivery without sacrificing quality, and you own the outcome regardless of how the code was produced
Proven ability to communicate technical concepts clearly to technical and non-technical audiences, including design documents and component documentation

Nice To Haves

Working knowledge of information-retrieval fundamentals (precision/recall trade-offs, hybrid search, ranking) and how low-level retrieval signals connect to downstream product quality
Experience building tools or APIs for agentic AI systems, including tool protocols such as MCP (Model Context Protocol) or similar
Background in legal technology, knowledge management, or other high-stakes information domains where source attribution, accuracy, and entitlement matter
Familiarity with search and retrieval engines (Vespa, OpenSearch, Elasticsearch) at the level of writing queries, designing indexes, or operating clusters
Experience integrating AI services — embedding APIs, re-ranking models, RAG-supported retrieval — with error handling and fallback strategies in production
Familiarity with API gateways (such as Apigee) and how a content API is secured, versioned, and exposed to consumers
Experience with agent and retrieval evaluation: deterministic eval suites and gold data gated in CI, red-teaming for prompt injection (e.g., Promptfoo), and tracing for LLM workflows (e.g., Langfuse)
Working understanding of infrastructure-as-code (Terraform) and container orchestration (ECS or Kubernetes), with the ability to configure and optimize CI/CD pipelines

Responsibilities

Design, develop, and own Content Tools end-to-end: the APIs and retrieval components that let agents query TR legal content through precise, well-documented, contract-driven interfaces, for example, jurisdiction-specific tools spanning legal content across multiple geographies
Build API-first: every tool ships with a versioned contract, schema, and ownership metadata so consuming teams and agent developers can use it self-service, without needing to ask you how it works
Design retrieval components for agentic use: respect explicit latency budgets across multi-hop chains, shape result payloads with token economy in mind, and build clean tool boundaries that give agent orchestrators predictable interfaces
Build and tune the retrieval and search behind the content tools in collaboration with our research scientists from TR Labs — improving relevance and ranking quality on our Vespa based search platform.
Practice verified spec-driven development: write the acceptance criteria and utilize evaluation datasets that define a tool’s behavior before you implement it — the eval is the spec — then build the eval suite, gold data, and CI gates that verify every change
Work in the team’s core stack — Python, FastAPI, and Pydantic on AWS, search in Vespa, with CI in GitHub Actions and observability in Datadog — building agent tools on the Claude Agent SDK and exposing them to agents over MCP (Model Context Protocol), and contributing to the shared libraries, generators, and patterns that help the whole team build tools correctly
Participate in the on-call rotation and take end-to-end ownership of incidents in your scope — triage, root-cause analysis, clear communication, and post-mortems that surface systemic fixes; you built it, you run it. Write and improve the runbooks for your areas.
Treat delivery friction as a bug: a slow deploy, a flaky test, or a painful local workflow is a problem worth fixing, because the team’s ability to ship constantly depends on every engineer pushing back on friction
Instrument your components with structured logging, metrics, and distributed tracing; for agentic retrieval paths, capture tool-invocation sequences and per-hop inputs so non-deterministic failures can be diagnosed and reproduced
Define and monitor SLOs for the tools you own, and treat fast, safe delivery as part of the job rather than a trade-off against reliability
Use AI-assisted development tools fluently across your workflow — generation, review, testing, debugging, — and model effective patterns for the team
Take technical direction for your tools with design-decision authority within your scope; break product and customer requirements into work the team can execute, and provide technical input on feasibility rather than only receiving tasks
Serve as the team’s go-to subject-matter expert for at least one content domain or tool family; provide domain-expert review on high-risk PRs and mentor more junior engineers through code review, pairing, and coaching
Work with our R&D division (TR Labs) and research scientists — alongside them, not receiving handoffs — to bring new retrieval capabilities into tools in a well-tested, observable, responsible way
Build working relationships with CoCounsel Engineering and product owners:
Capture decisions in writing through project-level ADRs, and communicate clearly to both technical and non-technical audiences
Champion ethical AI and secure, responsible engineering across your work