Lead​/Software Engineer, AI

Lead/Staff Software Engineer, AI

Are you ready to shape the future of AI‑driven tax automation? We’re looking for a Senior AI Engineer to build and scale intelligent workflows using modern LLM/ML tooling. Your work will directly impact how tax professionals automate complex, high‑stakes processes and access critical information.

As a Lead/Staff Software Engineer, you will own the design, implementation, and optimization of AI‑powered tax workflows—from data ingestion to production deployment—on our Tax and Accounting Professionals 1040 Scan product engineering team.

• Advance document intelligence:
  Build general‑izable models for tax documents using CV/NLP/LLMs and embeddings to move beyond fixed OCR templates with dynamic, context‑aware parsing.
• Boost auto‑verification & quality:
  Improve native PDF/textlayer matching, anomaly detection, and prior‑year‑aware checks to catch issues before human review; design human‑in‑the‑loop flows that preserve practitioner control.
• Scale the pipeline:
  Productionize low‑latency training/inference pipelines over millions of documents with robust observability, evaluation, and drift monitoring.
• Integrate LLMs and traditional ML models into robust, auditable workflows that support tax determination, document processing, and rules‑based automation.
• Build and maintain data pipelines, feature extraction, and preprocessing for tax‑relevant data (e.g., invoices, filings, transactional data).
• Develop and integrate RESTful APIs / microservices to expose AI capabilities to internal and external systems.
• Ensure solutions meet compliance, security, and auditability requirements typical of tax and regulated domains.
• Collaborate across Sure Prep engineering and other Thomson Reuters product & engineering teams leveraging TR’s AI platforms—including Materia and Additive—to deliver reliable, auditable AI in production.

• 5+ years in applied ML/AI (or 5+ with an advanced degree) delivering production systems in document AI/OCR/NLP or information extraction at scale.
• Strong Python and deep learning expertise (PyTorch/Tensor Flow), with experience in LLMs, embeddings/vector search, prompt design/evaluation, and statistical methods for quality and uncertainty.
• Cloud‑native ML ops experience: training/inference pipelines, feature/data stores, observability, cost/performance optimization.
• Own end‑to‑end lifecycle for AI features: design, implementation, testing, deployment, monitoring, and iteration.
• Optimize model/workflow inference for latency, throughput, and cost, including caching, batching, and model‑selection strategies.
• Implement monitoring, logging, tracing, and alerting for workflows and models in production (data drift, model performance, error rates).
• Debug and troubleshoot complex AI systems in production, including prompt failures, integration bugs, and data quality issues.
• Partner with Product, Tax SMEs, and other engineering teams to translate business/tax requirements into AI workflows.
• Collaborate with data scientists and ML engineers on model selection, evaluation, and experimentation within Reducto pipelines.
• Lead design reviews, perform code reviews, and mentor mid‑level/junior engineers on AI best practices and production‑grade coding.
• Contribute to internal standards, best practices, and templates for building AI workflows with Reducto.
• Built end‑to‑end AI features running in production, not just prototypes.
• Turn vague tax/business requirements into concrete, testable workflow designs.
• Care about robustness and correctness as much as cleverness—especially in domains where errors have financial or compliance impact.
• Comfortable reading tax/business rules and working with domain experts, even if not a tax specialist.

• Document AI background: layout parsing, table extraction, doctype classification, native PDF matching, and verification automation.
• Agentic/LLM orchestration experience with TR AI platforms such as Materia and Additive; retrieval‑augmented and tool‑use patterns; evaluation and guardrails.
• Modern development practices: hands‑on AI coding assistants (Claude Code, Cline, Git Hub Copilot) for rapid prototyping, code generation,…