Jordan Voves

Data Architect (2025–Present), Senior Data Engineer (2022–2025)

Joined as one of the founding members of a new AI team at Conexiom, a leading sales order automation platform. As the sole Data Engineer for nearly three years, I established Databricks as the company’s Data + AI foundation and architected a cutting-edge lakehouse in a greenfield Azure environment—now supporting multiple AI products, BI initiatives, real-time evaluation, and comprehensive operational monitoring. Promoted to Data Architect to scale the platform and guide cross-functional delivery.

Leadership

• Served as the primary technical interviewer for four hires and hiring manager for a Senior Data Engineer; now manage and mentor that role.
• Technical lead of an 8-person pod (AI Engineer, 3 SWE, FE, UI Designer, PM) delivering the Insights and Actions product.
• Provided data architecture leadership and strategic guidance as the most senior data engineering professional in the company.

Key Products

• Insights and Actions – Created and validated a new approach for generating automation insights from CSR correction data. Presented it through a Databricks App (Insights Lab) with Lakebase; strong reception led to business prioritization and production delivery.
• Navigator – Built the data foundation for a customer-facing app with KPIs and metrics giving customers real-time visibility into automation performance across 15 million+ trade documents annually.
• Express (AI Document Extraction) – Architected the data and evaluation layers underpinning Conexiom’s flagship AI extraction product, enabling >99.9 % uptime with full operational and cost transparency.

Context & Evaluation Frameworks

• Built a unified evaluation framework combining ground-truth metrics for extraction with LLM-as-judge for unlabeled tasks, enabling rapid iteration and confident production releases.
• Developed a context moat to enrich inference and continuously improve extraction accuracy:
  - Part Number Recommendation: Five-million-part catalog powered by vector search for context-aware matching when part numbers are missing.
  -Trading Partner Insights: Durable contextual assets built from user corrections, interactions, and historical performance.
  - Layout & Partner Detection: Embedded document layouts to accurately identify trading partners and document types.
  - Few-Shot Retrieval: Surfaced verified extractions from prior documents as high-quality examples during inference.

Platform Architecture

• Championed Databricks as the company-wide AI/data platform.
• Built and maintained 20+ production-grade Spark streaming/batch pipelines with automated CI/CD, alerting, and self-healing.
• Established Unity Catalog governance, evolving the lakehouse into a secure, production-grade platform supporting real-time monitoring and end-to-end observability across the AI system.
• Developed and maintained 10+ department-specific dashboards for finance, product, customer success, and engineering.
• Created Genie spaces for retrieval-augmented analytics and maintained a semantic layer (via Pydantic) to improve the effectiveness of Genie and AI/BI.

Sr Data Engineer (Jan 2022 - Jul 2022), Data Engineer (2018-2022)

• Led migration from legacy MSSQL workflows to a modern data lake architecture.
• Designed and implemented an Airflow-based orchestration platform, retiring dozens of legacy ETL jobs and enabling scalable production pipelines.
• Rebuilt large-scale aggregations (3B+ records) using PySpark and Presto, cutting runtimes from 48 hours to 20 minutes.
• Mentored junior engineers and data scientists on Python design patterns and AWS technologies (Kinesis, Glue, Lambda, ECS, Athena)