Senior Machine Learning System Builder

OPSWAT, a global leader in IT, OT, and ICS critical infrastructure cybersecurity, delivers an end-to-end platform that gives public and private sector organizations and enterprises the critical advantage needed to protect their complex networks, secure their devices, and ensure compliance. Over the last 20 years our commitment to innovative technology has earned the trust of more than 1,700 organizations, governments, and institutions globally, solidifying our role in protecting the world’s critical infrastructure and securing our way of life.The Position As a Senior Machine Learning Systems Builder you own detection capabilities end to end: from problem framing and data, through experimentation, to models running in production and the telemetry that proves they work for customers. We've had a strong presence in Veszprém for over a decade, and we're now expanding into Budapest, this role is based in our newly opening Budapest office, right at the start of that growth. You Will Have an Opportunity to Own detection capabilities from idea to customer impact across the AI/ML detection portfolio: threat similarity search (behavioral, code-structure, and static features), URL reputation, image-based phishing and brand-spoofing detection, web threat classification, and content classification Build and maintain the data pipelines your models depend on: sample sourcing and collection, ground truth and labeling workflows, feature extraction, and versioned training and evaluation datasets Design, train, fine-tune, and evaluate models, and see them through to production: versioned, runtime-portable artifacts (e.g., ONNX) consumed from the JVM-based backend, with input/output specifications, performance benchmarks, and documented limitations Build an automated model build and release pipeline, in the likes of SageMaker Pipelines: reproducible training runs, automated evaluation gates, and versioned artifact publishing, so that retraining and releasing a model is a routine operation rather than a project Own model quality in production, not just at release: telemetry feedback loops, false positive escalations, drift monitoring, and retraining cadence Design evaluation methodology for an adversarial, drifting domain: time-split validation, strict false positive ceilings, and robustness against evasion Automate the ML lifecycle with AI: use AI-assisted development daily, and build agentic automation for repetitive work such as labeling assistance, evaluation runs, regression testing, and reporting Set your own experimentation roadmap, prioritized by measurable customer-facing detection gains, and share what you learn openly across the team What We Are Looking for 3+ years applied ML experience across at least two of: text/content classification, computer vision, similarity search / embedding models, security or threat detection Evidence of end-to-end delivery: models you personally took from data to production, running in systems used by other teams or custom...