ML Engineer / Data Scientist - Core

ML Engineer Opportunity At Hilbert AI

Hilbert is building the ML systems that power demand intelligence for the world's largest consumer companies — recommendation engines, demand forecasting, customer lifecycle models, and activation systems that must work across wildly different retailers, data environments, and business contexts. This isn't single-tenant model building; it's designing configurable, production-grade ML systems that generalize across Fortune 500 enterprises and beloved consumer brands alike.

We're looking for an ML Engineer who understands B2C business problems deeply, builds models and pipelines that work with real-world data, and delivers systems that drive real growth outcomes — all with the ownership and urgency of a startup.

This is not a "receive a ticket, train a model, hand off a notebook" role. You'll own problems end-to-end — from framing through modeling through production deployment through impact — for enterprise customers where the stakes are real and the feedback loop is tight. If you understand why churn analysis matters differently for a grocery retailer versus a fashion marketplace, can build a recommendation system that works with sparse data and runs reliably in production, and can walk a customer through your causal analysis with clarity and conviction, we want to meet you.

Why Hilbert AI

Hilbert is building the demand intelligence platform used by world-class B2C leaders — including the world's largest retailer — to unlock compounding growth outcomes. We sit at the intersection of AI, data, and commercial activation for retail and e-commerce.

We're scaling fast with top-tier investors behind us. ML systems are the engine behind what we deliver to customers — which means every model you build, every pipeline you ship, every system you contribute to has direct, measurable impact on enterprise revenue. We're a small, talent-dense, low-ego team. We value ownership, speed, intellectual honesty, and shipping real impact.

The Role

You'll work directly with the founding team and alongside engineering, product, and GTM to build and improve the ML systems at the heart of Hilbert. You'll be hands-on every day — building models, designing pipelines, running experiments, interrogating data, and shipping to production. B2C is our world. The problems we solve — demand prediction, customer lifecycle, personalization, activation — require someone who understands these domains and can translate business context into modeling and engineering decisions. The environment is high-autonomy and high-ambiguity. Data is often messy, incomplete, or limited. You thrive in exactly those conditions.

Our Current Hurdles

These are the kinds of problems you'll be working on from day one.

• Multi-tenant ML systems that actually generalize — we serve enterprises with fundamentally different data shapes, catalog sizes, customer behaviors, and business constraints. The challenge is contributing to model architectures and pipelines that are configurable and adaptive across customers — not rebuilding bespoke systems for every account. You'll work on the abstractions that make this possible.

• Extracting real signal from messy, limited data — enterprise data is never clean and rarely complete. Cold-start problems, sparse interaction histories, inconsistent taxonomies, missing features — this is the norm, not the exception. You'll need to make pragmatic modeling choices that produce real value when the data fights back.

• Connecting model outputs to business actions — a recommendation score or a demand forecast is worthless if it doesn't change what an operator actually does. The challenge is closing the loop between ML outputs and real commercial decisions — activation, merchandising, retention — in a way that's measurable and defensible.

• Causal rigor in a world that wants quick answers — enterprise customers want to know why something is happening, not just what . The challenge is applying causal inference in a way that's rigorous but practical — knowing when an A/B test is sufficient, when you need difference-in-differences or synthetic controls, and when the honest answer is "we can't know yet."

What you'll do:

• Build and deploy ML models and pipelines that power core product capabilities: recommendation systems, search relevance, customer segmentation, demand forecasting, and activation optimization

• Contribute to configurable, multi-tenant model architectures that adapt across different customer contexts, data availability, and business requirements — not bespoke rebuilds for every use case

• Own your models through to production — working with engineering on serving, monitoring, and reliability, not just handing off prototypes

• Create meaningful models with the data that's actually available — not the data you wish you had. You extract signal from limited, noisy, or sparse datasets and reach for the right level of complexity

• Design and run rigorous A/B tests — including understanding when A/B testing is insufficient and causal inference methods are required

• Apply causal reasoning rigorously — you know the difference between correlation and causation, you surface true drivers, and you flag when others confuse the two

• Deliver analyses that drive decisions — you connect model outputs to business outcomes and communicate them with clarity to founders, teammates, and customers

• Think in systems. You don't build isolated models — you understand how recommendation, segmentation, scoring, and activation interact with each other and design your work to fit within the broader system

• Move fast — prototype, validate, ship, iterate. You're comfortable with imperfect information and evolving requirements

Who You Are

We care about how you think about problems, how you connect models to business impact, and how you operate when things are ambiguous.

The profile:

• You're an ML engineer who ships to production. You write clean, testable Python. You care about how your models are served, monitored, and maintained — not just how they perform offline. Your work doesn't end at a notebook; it ends when the system is running and delivering value.

• You're a product-minded ML engineer. You understand that a model with great offline metrics is useless if it doesn't move the customer's business. You frame every modeling decision in terms of the outcome it enables — and you push back when a technically elegant approach doesn't serve the actual problem.

• You have strong B2C business knowledge. You understand the problems consumer businesses actually face — customer acquisition vs. retention economics, lifecycle dynamics, basket composition, churn drivers, promotional cannibalization, channel attribution, demand elasticity. This knowledge informs how you frame problems and design models.

• You're a systems thinker. You understand how models, data flows, customer behavior, and business outcomes connect. You don't optimize one metric in a vacuum — you consider second-order effects and how your work fits the bigger picture.

• You've built recommendation, search, and/or customer-based ML systems — collaborative filtering, content-based methods, ranking systems, segmentation, propensity modeling. You understand when each applies and why.

• You know how to build for configurability. You've worked on or contributed to model architectures and pipelines that flex across multiple customers, segments, or contexts — not rigid, single-purpose implementations.

• You create value from limited data. You make pragmatic modeling choices when data is sparse, noisy, or cold-start. You know when a simpler approach beats a complex one and aren't seduced by unnecessary sophistication.

• You're rigorous about causality. You understand causal inference methods — difference-in-differences, instrumental variables, propensity scoring, synthetic controls — and know when to apply them. You design A/B tests properly and understand their limitations.

• You communicate with clarity and conviction. You can present an analysis to a non-technical audience and make it land. You can write a one-pager that changes a decision. You explain your reasoning, not just your results. Communication is not a nice-to-have here — it's core to the role.

• You take ownership. You don't wait for someone to define the problem perfectly. You dig in, frame it, propose an approach, and ship it. If something breaks or underperforms, you treat it as your problem.

• You thrive in ambiguity. Problem definitions shift. Data availability surprises you. Requirements evolve. You