Senior Machine Learning Engineer

About Leon Capital Group

Leon Capital Group (LCG) is a multi-billion-dollar holding company that owns and operates businesses across healthcare, real estate, and financial services. We found, acquire, and scale companies for the long term, backing them with shared capital, talent, and infrastructure while giving each the room to operate and grow. This role sits inside AMP, our medical aesthetics business. You are employed by AMP, report to the Chief Innovation Officer of LCG, and carry a dotted line to AMP leadership.

About the Role

AMP runs roughly 75 clinics across multiple brands, and licensed provider time is the scarce resource the P&L turns on. The data already exists — schedules, patient and treatment history, marketing engagement, consumption. And we want to amplify its operational impact.

This role is the machine learning anchor for that work: the engineer who owns the intelligence layer AMP runs on and stays with it as it compounds. The core of it is applied predictive modeling, extending into agentic systems that run in production against live operational tools rather than demos. We want someone energized by owning an outcome, not a metric — who measures success in utilization, labor cost, conversion, and retention, and treats a licensed provider as the decision-maker their model advises.

On scope: you are the only machine learning hire at AMP, and there is no ML team above or beside you. You set the approach, choose the models, and decide what ships. A platform engineer builds the warehouse and marts alongside you through the first months and then shifts to other parts of the portfolio, so expect to own the feature and data layer your models depend on rather than inherit a team to maintain it. This is a role for someone who wants that ownership, not a lane inside a larger team.

Key Responsibilities

• Forecast provider demand across roughly 75 clinics and match staffing to it, raising provider and staff utilization and lowering labor cost as a share of revenue. This is the first priority.
• Design and ship agentic systems that act, not just predict, running against live operational systems rather than demos. Procurement and scheduling are the first candidates.
• Extend the predictive layer into patient acquisition and retention (scoring, targeting, and treatment recommendation) to grow new-patient volume, conversion, and retention.
• Own the full model lifecycle (feature engineering, training-set construction, evaluation, calibration, monitoring, and retraining) plus the evaluation discipline that keeps agents reliable in production.
• Work under clinical and regulatory constraints throughout — handling PHI with the secure practices HIPAA requires, and designing human-in-the-loop workflows where a licensed provider, not the model, makes the call.

Qualifications

Required:

• 5+ years in applied machine learning, ideally as an early or anchor ML hire who owned models end to end rather than one stage of a large team.
• Strong Python and the applied ML stack: pandas, scikit-learn, and gradient boosting (XGBoost, LightGBM) as the working default, with PyTorch or TensorFlow where a problem genuinely calls for it.
• Applied ML and statistical modeling depth: feature engineering, training-set construction, evaluation, calibration, interpretability, and the judgment to reason about model risk. You orchestrate and apply models; you do not need to publish research.
• Time-series and demand forecasting depth — seasonality, cyclicality, and multi-site heterogeneity — and the judgment to know when a simple model beats a sophisticated one.
• A track record shipping predictive models into production decisions — demand or utilization forecasting, propensity and lead scoring, recommendation, or similar.
• Strong SQL and comfort working directly on curated warehouse marts (Azure SQL, dbt), plus the flexibility to reason across a multi-platform estate. You can shape features from the data layer without a separate team.
• Familiarity with the tooling that keeps models in production; experiment tracking and a model registry (MLflow or similar), pipeline orchestration (Airflow, Dagster, or similar), containerization, and monitoring for drift and retraining.
• Hands-on experience building agentic, tool-using LLM systems that run against real systems rather than demos — orchestration (e.g. LangGraph), retrieval-augmented generation, tool and function calling, and disciplined prompt and evaluation practice.
• Judgment about model trust in a setting where a licensed provider acts on the output — calibration and thresholding you can defend, and the ability to explain a model's behavior, assumptions, and limits to a clinician or an operator.
• The ability to work effectively across implementation partners, consultants, and platform vendors without becoming blocked by them.
• Excellent judgment under ambiguity and the ability to prioritize and ship without close oversight.
• A bias toward measurable business outcomes: you tie models to utilization, labor cost, conversion, and retention, not to offline metrics alone.

Preferred:

• Healthcare, medical aesthetics, or multi-site provider-operations experience, where you built to clinical and compliance constraints.
• Hands-on experience with workforce or labor optimization, scheduling, or supply-and-demand matching.
• Inventory, procurement, or supply-chain optimization.
• Recommendation or next-best-action systems in a customer or patient setting.
• Experience unifying data across multiple platforms or post-acquisition environments.
• You have been the person a business's models depended on, at a place small enough that there was no one else to do it.