Senior Applied AI Engineer, Document Intelligence

Headquarters: New Jersey, USA URL: https://vertican.com About the role For 40 years Vertican has buildt software that runs law firms. One year ago we created a new modern AI platform for document automation (processing/extraction/redacting) which our customers have been adopting faster than any product we've ever had. Our suite of AI-enabled products save customers thousands of hours each week by processing high volumes of documents with precision and speed.The AI team is growing and we'd like you to join. We are hiring an engineer to own the document processing pipelines end to end: design, benchmarking, experimenting with new architectures and tools, threshold required for deploying a new pipeline, etc. We believe it's important to build our own tooling to help us run, improve, and measure our applications. Therefore, the ideal candidate thinks outside the box on how to not just build the product/pipeline, but also the tooling needed around it. You will be the first dedicated ML hire on the product taking over from the engineering manager who build the current solution and will work directly with the founding engineering team. This is a senior, hands-on individual-contributor role, but we are small team and you'll be involved in design, planning, roadmapping, and architectural decisions. What you will do Design the pipelines. Own the architecture of the extraction-to-posting pipelines: page classification, field extraction, candidate retrieval, claim matching, confidence gating and the hand-off to human review, etc. Decide where rules end and models begin, and what each stage must guarantee to the next, what the separation of concerns are, etc. We have different pipelines for different document types and expect that at any given point in time we may have new ones emerge and may consolidate others. Evaluate new systems before we adopt them. Run structured trials of candidate components and recommend what to keep. Examples include probabilistic record linkage (Splink), Azure AI Content Understanding and Document Intelligence, Azure OpenAI and Foundry-hosted models, and LLM agents with tools. Similarly, test new models with the existing pipeline as they are relased from the frontier labs. Build the evaluation process. You'll own the process of evaluating false positives, false negatives, errors of ommission, etc, that we ultimately use to determine if we should upgrade the pipeline. Ship and operate. Take components from trial to production in our python workers on the cloud, instrument them, watch cost and quota, and diagnose the cases reviewers send back. Set direction. Work with leadereship on the planing for accuracy, cost improvement, and overall product improvements. What we are looking for Required • 6+ years building ML or data-intensive systems, at least 3 of them owning a production pipeline whose output people acted on. • Solid backend engineering understanding and ideally cloud architecture understanding. • Ideally hands-on experience with document understanding: OCR or layout models, field extraction from scanned PDFs, or intelligent document processing products. • Experience with entity resolution or record linkage at scale (Splink, Dedupe, Zingg, or your own blocking-and-scoring system) against databases of hundreds of thousands of rows. • You have designed an evaluation from scratch: built a labeled ground-truth set, chosen the metrics, and defended a threshold to a non-technical stakeholder. • Production experience with LLM APIs, including structured output, tool calling and cost or rate-limit management; you can say when an LLM is the wrong tool. • Strong Python and SQL. Comfortable reading Postgres query plans and writing the raw query when the ORM gets in the way. • Engineering/Scientist