Senior Machine Learning Engineer, Personalization & Decision Science

Description

The company runs an AI grocery-intelligence platform that helps people plan meals and shop, serving millions of consumers through its own app and a white-label enterprise suite.

The company is profitable, handles billions of dollars in transactions for large US retailers, and its mobile apps reach millions of people.

The team is hiring a senior machine learning engineer for personalization and decision science to build the models that choose which meals, recipes, products, substitutions, replenishment items and full baskets to recommend.

The hard part goes beyond predicting what a household likes. A plan of five chicken recipes may match taste yet still be repetitive, so the models must also weigh servings, budget, nutrition, diet, cooking time, variety, package sizes, ingredient reuse, pantry, inventory and real preferences together.

This is a senior individual contributor role with direct ownership of production systems. You will turn the personalization strategy and system goals into models, constraints, optimization methods and evaluation plans, then run and improve them after launch.

The stack includes Python and Django on the backend, JavaScript and React Native on the frontend, and GraphQL with real-time streaming over WebSockets. The data covers recipes, ingredients, products, retailer catalogs, transactions, carts, prices, promotions, inventory, nutrition, preferences and pantry signals.

Responsibilities

• Design and ship multi-stage recommendation systems covering candidate generation, retrieval, filtering, scoring, ranking, reranking and final selection for recipes, products, meals, collections, substitutions and replenishment items.
• Learn individual and shared household tastes from preferences, orders, recipe use, cart edits, substitutions, removals, repeat buys and chat feedback.
• Separate durable taste from one-off context such as guests, holidays, parties, work lunches, children's food or a single recipe.
• Treat meal and grocery planning as a constrained multi-objective problem that balances household value against budget, servings, nutrition, allergens, convenience, variety, reuse, package sizes, pantry carryover, leftovers and availability.
• Forecast recurring needs from timing, quantity, household size, seasonality, cadence and uncertainty. Rank replacements by functional fit, recipe context, diet, brand, price, package size, stock and likely acceptance.
• Personalize for new households, members, recipes, products and sparse users. Map canonical foods to store SKUs and adapt to assortment, prices, promotions, stockouts, regional availability and catalog changes.
• Evaluate whole baskets, not single items, through constraint violations, diversity, calibration, coherence, waste, reuse, acceptance, cart edit distance, substitution acceptance, order behavior, repeat use and retention, using offline replay and online experiments.
• Own pipelines, batch and low-latency inference, caching, versioning, monitoring, drift detection, fallbacks and rollback. Expose candidates, scores, confidence, constraints, reason codes and alternative plans through services that apps and AI agents can consume safely.

Requirements

• At least 5 years in machine learning engineering, applied science, recommender systems, operations research or a related quantitative field.
• Shipped a significant recommendation, ranking, forecasting, personalization or optimization system to production and owned its performance after launch.
• Deep knowledge of recommenders, retrieval, ranking, embeddings or user modeling, plus solid working knowledge of forecasting, mathematical optimization or other decision-science methods.
• Familiar with collaborative filtering, content-based models, two-tower retrieval, sequence models, learning to rank, reranking, calibration, exploration and contextual bandits, and able to tell when a simpler method is the better choice.
• Highly proficient in Python and SQL, writing maintainable, well-tested production code rather than only research notebooks.
• Understand implicit feedback: exposure bias, negative sampling, delayed outcomes, sparse labels, popularity effects, feedback loops, and the gap between not chosen and disliked.
• Able to use experiments and causal reasoning to separate real gains from noise, leakage, bias or misleading offline results.
• Able to run the full production ML lifecycle, including feature freshness, training-serving consistency, latency, caching, monitoring, data and model drift, graceful degradation and retraining.
• Communicate clearly with engineers, product leaders, culinary stakeholders and founders, and care about what makes a meal complete, a week varied, a substitution valid and a cart believable.
• Bonus: experience in grocery, meal kits, food delivery, retail, e-commerce, marketplaces, subscriptions or another constrained recommendation setting.
• Bonus: experience with consumption or repeat-purchase forecasting, survival analysis, mathematical programming, constraint optimization or real-time recommendation systems.
• Bonus: a master's or PhD in computer science, operations research, statistics, applied mathematics, industrial engineering, economics or a related field.

Conditions

The role is based in Austin, TX, with work-from-home flexibility.

Benefits

• Pay is competitive and equity is meaningful.
• Collaborate closely with the founders on a tight-knit team.
• Build products that improve daily life for millions of people.
• A high-intensity, high-ownership setting built for people who like to build.