Senior Full-Stack Engineer (AI Workflow Systems) - Part-Time - Overlap with US timezone

Recap Technologies is building AI-native workflow software for financial consulting and advisory teams. The product helps consultants handle messy, high-stakes financial data — including bank statements, GL exports, Excel files, PDFs, and fragmented records — by transforming them into structured, auditable workflows and client-ready deliverables.

The immediate focus is an AI-powered engagement workflow platform for tasks like document ingestion, transaction reconciliation, variance analysis, entity-level reporting, and review workflows. This is not a chatbot product; the system must reason across inconsistent financial data, maintain auditability, explain confidence levels, and support human review.

One core use case involves multi-entity transaction reconciliation. The platform ingests financial documents from multiple sources, normalizes inconsistent data, applies deterministic and AI-assisted matching logic, flags discrepancies and suspicious transactions, generates review queues with audit trails, and learns from reviewer corrections over time. Accuracy, explainability, and reliability matter more than flashy AI demos.

The company is led by a founder with deep financial advisory and enterprise sales experience, building around real customer pain points from consulting firms handling complex client data. The role starts as a senior contract engagement with a strong path toward full-time technical leadership ownership.

Current stack includes:

• Frontend: Next.js / React
• Backend: Python and/or Node.js APIs
• Database: Postgres
• AI layer: Claude / Anthropic today, with future Azure OpenAI or Vertex AI usage
• Workflow tools: Cursor, Claude Code, GitHub

The company is moving toward an enterprise-grade cloud architecture focused on security, auditability, and scalability. Desired experience includes:

• Azure (preferred) or GCP cloud architecture
• Secure file ingestion/storage
• Managed relational databases
• Private networking and IAM/RBAC
• Encryption, secrets management, audit logging
• Event-driven workflows, queues, workers, and infrastructure-as-code
• Multi-tenant SaaS security practices

The engineer will help build production-grade AI workflow systems including:

• Document ingestion and extraction from PDFs, Excel, CSVs, and financial documents
• Data normalization and validation pipelines
• Reconciliation and exception-detection logic
• Human-in-the-loop AI workflows
• Secure review interfaces, dashboards, APIs, and background jobs
• Evaluation frameworks for AI extraction quality, latency, and reliability

Required experience:

• 5+ years in software engineering
• Strong full-stack experience (React/Next.js + Python or Node.js)
• Strong Python data engineering skills
• Production LLM experience beyond simple chat interfaces
• Experience with document extraction, structured outputs, RAG, agentic workflows, or AI-assisted data processing
• Strong Postgres/data modeling knowledge
• Background in asynchronous processing and file ingestion workflows
• Cloud platform experience (Azure preferred)
• Security fundamentals (IAM, encryption, auditability, secure storage)
• Ability to work independently and turn ambiguous workflows into practical products

Strongly preferred:

• Experience in finance, accounting, audit, compliance, or consulting workflows
• Transaction reconciliation or financial data QA systems
• Human-review AI systems with confidence scoring and review queues
• Multi-tenant SaaS products handling sensitive data
• Azure OpenAI, Vertex AI, OCR/document AI tooling
• LLM evaluation and regression testing frameworks
• Enterprise workflow software experience

The company values pragmatic builders over theoretical AI researchers. They are looking for someone comfortable with messy enterprise data, secure systems, and practical product delivery — not someone focused only on chatbot demos or over-engineered abstractions.

Example paid trial project:
Build a reconciliation prototype that ingests bank statements and GL exports, normalizes data, matches transactions using deterministic and fuzzy logic, flags exceptions, assigns confidence levels, and presents results in a review UI with audit trails.

Work style:

• Async-first with 1–2 live sessions per week
• Founder provides customer/domain context
• Engineer owns architecture and implementation decisions
• Small weekly milestones over large speculative builds
• Strong focus on real customer workflows and practical execution
• Potential long-term path into technical leadership and ownership of the platform architecture