ML / AI Engineer

Role overview Develop the AI capabilities behind a document-compliance workflow: incoming files are recognized, classified, converted into structured fields, checked against rules and prior records, and surfaced for human attention. You’ll guide how this layer evolves, from prompt design and evaluation to retrieval-based validation and predictive models. The work includes assessing model quality on domain-specific questions, such as whether a certificate’s reported measurement is consistent with a supplier’s history.

Responsibilities - Design and evaluate prompts and model behavior for document classification, field extraction, and chat-based workflows. - Decide when a task is better addressed through prompting or fine-tuning, using evaluation results to guide changes. - Develop validation using retrieval, embeddings, and historical records to check extracted information. - Move a fraud-detection capability from a mock implementation toward a deployed model. - Build and maintain evaluation approaches for renewal risk, recall exposure, and document-level consistency checks. - Use mock implementations at service boundaries to support development and CI without relying on live model quotas.

Requirements - Experience building or evaluating applied AI or machine-learning systems. - Ability to design meaningful evaluations for extraction and validation tasks, including comparisons with rules and historical data. - Familiarity with prompt engineering and the trade-offs between prompting and fine-tuning. - Ability to work with retrieval-based methods, embeddings, and structured document data. - Sound judgment about model quality, testing, and the limits of automated decisions.

Nice to have - Experience with Google Document AI, Vertex AI, Gemini models, or SageMaker deployments. - Background in OCR, retrieval-augmented generation, predictive risk scoring, or fraud detection.