AI / Backend Engineer (Automation & AI Platform)

About the RoleWe are building a production-grade, AI-enabled tax compliance platform to support the preparation of 5,000+ tax returns annually and enhance our existing workflows through structured automation and AI-assisted processing.This platform is designed to support and augment our tax teams, reducing time spent on repetitive data preparation while enabling greater focus on review, judgment, exception handling, and client advisory work.By improving efficiency and consistency across processes, the system will empower both our local and offshore teams to deliver higher-quality outcomes and scale effectively as volumes grow.This is a high-impact engineering role focused on building an end-to-end AI-assisted system that will play a key role in modernizing how our firm operates, while keeping human expertise at the core of decision-making.You will take end-to-end technical ownership of a business-critical system, leading design and delivery from architecture through to production deployment. You will work closely with tax and domain experts to translate regulatory and business requirements into scalable, production-grade system logic.Note: This is a production-grade system build (not a prototype or chatbot use case). The system must be reliable, auditable, and designed to work alongside professional reviewers in a controlled, compliance-driven environment.What You Will BuildThe system is designed around review-first, exception-based workflows, where routine, standardized cases are pre-prepared to a high level of accuracy, allowing professionals to focus on validation, exceptions, and advisory judgment.Key capabilities include:Document IngestionAutomatic retrieval of tax-related documents from document management systemsSupport for documents such as income statements, payroll records, contribution statements, and supporting documentsNo manual uploading requiredAI Data ExtractionUse LLM-based workflows to extract, validate, and structure financial dataConvert unstructured documents into structured datasetsReconciliation EngineCross-check data across multiple sourcesDetect inconsistencies, missing elements, and anomaliesAI-Assisted Case PreparationGenerate structured, review-ready cases with flagged issues and summariesProvide draft tax positions for reviewer validationProduce clear, plain-English summaries for efficient reviewFinal tax calculations will be handled by a deterministic, rule-based engine. AI is used to support extraction, validation, anomaly detection, and explanation—not to replace professional judgment.Workflow AutomationPre-process standard casesHighlight and escalate exceptionsEnable structured review workflows and trackingKey ResponsibilitiesAI & Backend DevelopmentBuild multi-step LLM workflows (extraction, validation, reconciliation, summary)Design structured data schemasImplement prompt pipelines with validation layersDevelop confidence scoring and fallback mechanismsIntegration & Data PipelinesIntegrate with document management systemsBuild ingestion and classification pipelinesProcess PDFs, scans, and structured inputsSystem DevelopmentBuild backend services (Python, FastAPI preferred)Develop internal dashboards for tracking, issue management, and workflowsTax Logic IntegrationBuild or integrate deterministic tax rulesEnsure outputs are auditable and structuredSecurity & ComplianceEnsure data protection complianceImplement role-based access controlMaintain audit trailsRequirementsMust-Have3–6 years of backend or AI engineering experienceStrong Python experience (FastAPI preferred)Experience building multi-step LLM pipelines with validation layers in productionExperience with document processing and structured data extractionStrong understanding of APIs, JSON, and data flowsAbility to design end-to-end data pipelinesGood to HaveExperience with financial, tax, or payroll dataExperience with workflow automation systemsFamiliarity with document systems (e.g. M-Files, SharePoint, S3)Experience with PDF parsing toolsDeliverablesMVP (10–12 weeks)End-to-end flow: ingestion, extraction, structured dataset, basic summariesProduction-ready with validation, logging, and error handlingFull System (6 months)Reconciliation engineException workflowsReviewer dashboardAudit and logging framework