Forward Deployed Engineer - Data Scientist

About Us Viaduct builds production-grade AI systems for the world's manufacturers: not pilots, not proofs of concept. The Internal Efficiencies team is deploying real AI use cases inside one of Japan's largest manufacturers - a modern cloud and AI stack, shipped all the way down to a plant floor running robots, lasers, and IoT gear most software teams never get near. We're a new team building fast. Viaduct already has a sophisticated, battle-tested platform - AWS landing zone, Terraform, AKS, Postgres, ClickHouse, Argo, observability - with genuinely great bones. But that platform was built to host production SaaS applications, and our team's needs don't look like that. We need our own environment, built on the lessons of what already exists, but shaped around how we actually work: fast iteration, new tooling, production use cases shipping constantly. Who You Are You're motivated by real-world problems, not toy ones, and you want to see your analytical work actually change how a real business or manufacturing process runs. No manufacturing background required, we'll teach you the domain, though if you've already got industrial or manufacturing experience, that's a genuine plus. You're also naturally curious about the business problem behind the data. You think beyond the model itself, starting with the decision we're trying to improve, the drivers of the problem, the size of the opportunity, and how we will measure whether we've actually created value. About the Role You'll be one of a small team of high performers anchoring the analytical underpinnings of our work end to end: identifying opportunities, framing ambiguous problems, calculating their impact, and building the advanced LLM-powered pipelines and agentic systems that automate and optimize both business and manufacturing processes, from plant-floor operations to demand forecasting, for our manufacturing partners. Think of it as internal consulting with real deployment authority, you're not handing off a deck, you're shipping the thing you scoped. A big part of the role is translating messy business problems into structured analytical problems. You'll work with stakeholders to understand the current process, diagnose root causes, develop hypotheses, quantify potential value, and determine where data science or AI can make a meaningful difference. You'll then take that problem from analysis through solution design, development, deployment, and measurement of impact. We're looking for a Data Scientist with real experience delivering production use cases, not a first DS role. What You'll Do Day to day, you'll live in Python and SQL, building and shipping models with tools like scikit-learn, XGBoost, and increasingly agentic workflows, against datasets that don't always fit on one machine (AWS, Argo Workflows, Clickhouse). You'll move seamlessly between structured problem solving, quantitative analysis, and hands-on engineering. You're as comfortable presenting to a business stakeholder as to another engineer, and a model isn't done in your book until it's deployed and maintained, not just validated in a notebook. What We're Looking For Real experience shipping ML models to production, not just notebooks and research Strong Python and SQL skills, hands-on with scikit-learn, XGBoost, or similar ML libraries Some hands-on experience with PyTorch and other deep learning technologies Strong structured problem-solving skills: able to break ambiguous business problems into hypotheses, analyses, and actionable recommendations Business acumen and comfort thinking about value: able to quantify the potential impact of an opportunity and connect analytical outputs to business decisions and outcomes Ability to frame problems and communicate a clear "so what", not just what the data says, but what should be done about it Comfortable working directly with customers and explaining technical work to non-technical audiences Experience working with business stakeholders to understand processes, identify pain points, and translate them into analytical or technical requirements Experience with distributed or cloud data tools (AWS, Clickhouse, Argo Workflows, or similar) Experience deploying and maintaining production data pipelines or DS-backed product features You work AI-native: you use AI tools to accelerate your own work, not just build AI for others Nice to have: experience building or fine-tuning LLMs, and developing agentic solutions and creating impact through working in an analytics consulting, or other client-facing problem-solving environment. Your First 90 Days In your first weeks, you're in the data, not in a generic onboarding deck. You'll meet the six-person team, including our resident ex-McKinsey/QuantumBlack data scientist, and get your hands on real data from one of our engagements, that might be sensor and maintenance data off a manufacturing plant floor, or demand data feeding a forecasting model. You don't need to know the domain going in, you'll pick it up fast by sitting in on partner conversations and hearing directly what they need. By the end of your first month, you've shipped something real, not a proof of concept. You'll have framed a business problem, diagnosed the opportunity, quantified the potential impact, scoped and proposed an improvement, decided whether to build it fresh or build on our existing internal platform, gotten it into our ML or agentic pipelines, and presented it at a team meeting. A few months in, you're driving your own slice of the roadmap. You're deploying production-grade DS products, pairing with engineers to sharpen our tooling, and presenting your models and their real business impact directly to our partner's stakeholders, not just to the team. The Team We're remote and distributed, and we get together in person about once a month. Right now we're about eight people, growing through the end of the year, with a deliberate mix of consulting and hands-on backgrounds, people who move across business, strategy, and engineering rather than staying in one lane. You'd be joining a team that already includes a data scientist with a McKinsey/QuantumBlack background, so the bar for rigor and business fluency is real. Security and Privacy Responsibilities Member of the CSIRT Team Follow our policy and procedure documents related to security and privacy Follow the guidelines in the Employee Handbook Participate in new hire and annual training for security and privacy Treat data security and privacy as one of your primary job responsibilities Report Security Incidents you discover as bugs Get approval from the Security Team before adding new 3rd party software to our codebase Explicitly consider security implications when doing PR reviews