ML Engineer

About the role

ApprovalMax is a fast‑growing B2B SaaS company that automates approval workflows and financial controls. The Capture product processes hundreds of thousands of invoices, bills and purchase orders each month through an OCR pipeline that feeds accounting systems. As an ML Engineer you will own the effort to raise the zero‑touch rate – the share of documents that require no manual correction – by improving entity matching, pipeline logic and OCR extraction.

Key responsibilities

• Investigate accuracy failures at scale, query production datasets and identify statistical patterns behind document errors.
• Own and evolve the accuracy measurement framework, defining uplift expectations and post‑deployment monitoring.
• Develop and ship ML services for entity matching using embedding‑based similarity search and recommender techniques.
• Design rule‑based fixes for pipeline‑logic errors and validate them on production data.
• Build an OCR correction layer – leveraging LLMs, alternative OCR engines or HuggingFace models – and rigorously evaluate its impact.
• Collaborate with remote teams across the UK, Serbia and Moldova to prioritize the backlog and deliver measurable improvements.

Required profile

• Strong analytical mindset with experience in forensic data investigation.
• Proven ability to take ML models from prototype to production in a SaaS environment.
• Comfortable working independently and driving end‑to‑end projects.
• Familiarity with financial document processing is a plus.

Required skills

• Optical character recognition (OCR) technologies.
• Large language models (LLM) with appropriate guardrails.
• Embedding‑based similarity search.
• Recommender system design.
• Rule‑based pipeline development.
• Experience with HuggingFace model ecosystem.

What we offer

• Fully remote work for candidates based in the UK, Serbia or Moldova.
• Opportunity to own a critical ML product in a rapidly scaling company.
• Collaborative global team and competitive compensation.