AI data architect - health and wellness

Title: AI & Data Architect – Personal Health and Behavioral Intelligence About the Role We are building a consumer preventive-health and behavior platform that combines authorized data from Apple HealthKit and Android Health Connect with camera-based food, ingredient and facial scans, wearable IoT signals, user feedback, conversational information and in-app quests. We are looking for a hands-on AI & Data Architect who can design and build the first version of our Personal Health Model and the data platform supporting it. This is not a pure research, management or advisory position. The successful candidate must be comfortable defining the architecture, writing production code, building data pipelines, testing practical models and working directly with our mobile, product and external IoT teams. The immediate goal is not to create a medically diagnostic AI system. It is to organize incomplete and multimodal consumer wellness data into personal baselines, structured life episodes, testable personal observations and relevant behavioral quests. What You Will Own - Overall AI, data and machine-learning architecture - Canonical data model for health, camera, IoT, behavioral and user-reported data - Personal Health Model and longitudinal user profile - Data aggregation, normalization, provenance and confidence methods - Initial model-training and evaluation pipelines - Connection between personal-health intelligence and quest selection - Technical direction of external AI, IoT and data vendors - Internal ownership of source code, schemas, models and technical documentation Key Responsibilities Data and Platform - Design and implement ingestion pipelines for Apple HealthKit, Android Health Connect, camera-scan outputs, IoT events, user feedback and product interactions. - Normalize units, timestamps, time zones, device sources and duplicate records. - Handle missing data, revoked permissions, inconsistent measurements and differently sampled data. - Design storage for structured health records, time-series signals, images, audio, transcripts and model outputs. - Preserve data provenance so each derived result can be traced to its original input, processing version and model version. - Implement appropriate privacy, retention, export and deletion controls. - Build practical APIs and services for the mobile app and quest system. Personal Health and Multimodal AI - Develop the first usable Personal Health Model using a combination of rules, statistics, machine learning and foundation models. - Convert raw observations into structured personal episodes. - Build personal baselines and identify meaningful changes from an individual’s normal pattern. - Combine camera-derived observations with wearable, health-platform and user-reported data. - Attach confidence and uncertainty values to observations and derived associations. - Create cold-start methods for users with limited data. - Design labeling guidelines and datasets for future model training. - Establish train, validation and test procedures that prevent user and session leakage. - Decide which functions should use traditional ML, generative AI, deterministic rules or human confirmation. Quest and Behavior Intelligence - Translate personal observations into safe, small and measurable behavioral quests. - Design initial quest eligibility, ranking, timing and suppression logic. - Use quest completion and user feedback to update the Personal Health Model. - Establish methods for measuring engagement separately from meaningful behavior change. - Prepare the architecture for later use of recommendation models, contextual bandits or adaptive interventions without overengineering the first release. Camera and IoT Collaboration - Work with existing or third-party computer-vision models for food, ingredient, packaging and facial-image analysis. - Define the structured outputs, confidence requirements and APIs expected from camera models. - Define raw-data formats and synchronization requirements for external IoT teams. - Evaluate whether vendor models and sensor outputs are technically credible. - Integrate external models and services without surrendering control of our core data and architecture. What We Expect You to Build First During the first stage, you will be expected to deliver: 1. An audit of the current application, databases, health integrations, camera features and quest system. 2. A canonical observation and event schema. 3. A working pipeline for selected HealthKit, Health Connect and camera-derived data. 4. A first version of the longitudinal Personal Health Model. 5. A rules-first personal baseline and change-detection engine. 6. A structured personal-episode generator. 7. A basic quest recommendation and feedback loop. 8. A practical dataset, labeling and model-training roadmap. 9. Technical requirements for the external IoT team. 10. A staged plan identifying what should be built internally, purchased, outsourced or deferred. Required Qualifications - Strong hands-on experience in Python, backend systems, databases and production AI/ML. - Experience building data pipelines and deploying models or AI services. - Understanding of time-series, multimodal or longitudinal data. - Ability to work with structured data, text, images and sensor-derived features. - Practical understanding of modern ML, LLMs, embeddings, evaluation and MLOps. - Ability to design systems that handle noisy, incomplete and asynchronous data. - Strong product judgment and willingness to begin with a simple measurable baseline. - Ability to communicate clearly with founders, mobile engineers, product teams and external hardware vendors. - Willingness to write code and personally deliver the first production system. Preferred, Not Mandatory - Apple HealthKit, Android Health Connect, FHIR or wearable-data experience - Digital health, nutrition, consumer wellness or behavior-change products - Computer vision or vision-language models - Recommender systems and personalization - Audio, physiological-signal or IoT data - Privacy-sensitive consumer-data systems - Startup or zero-to-one product experience We do not expect one candidate to be a specialist in every listed field. We are looking for someone with a strong AI and data-engineering foundation, broad technical literacy, sound judgment and the ability to learn or engage specialists when necessary. Application Questions 1. Describe one AI product for which you personally designed the architecture and wrote production code. 2. How would you combine HealthKit or Health Connect data, camera-derived observations, IoT events and user feedback into a longitudinal personal model? 3. What would you build with deterministic rules first, and what would you train as a model later? 4. How would you make the system useful before we have a large proprietary dataset? 5. How would you represent missing data, conflicting measurements and uncertainty? 6. How would you connect a personal observation to a behavioral quest and determine whether the quest worked? 7. Which parts of this scope would you personally build, and where would you recommend using an external specialist or existing model? 8. What would you aim to deliver during your first 30, 60 and 90 days?