Senior AI/LLM Architect & Developer — Conversational Digital Health Platform

Senior AI/LLM Architect & Developer — Conversational Digital Health Platform We are developing a serious AI-powered digital health platform, initially focused on helping people quit smoking and nicotine. The treatment methodology and clinical program are being developed separately by an experienced smoking-cessation professional. (We are not looking for someone to develop the treatment itself.) We need an experienced (AI/LLM architect and developer, or a small specialized AI team), who can translate our structured treatment methodology into an intelligent, personalized conversational system. This is intended to become a long-term commercial platform — ( not simply a chatbot or ChatGPT wrapper.) What the system needs to do The AI counsellor will guide clients through a structured treatment program consisting of multiple sessions over several months. The system should eventually support: * Natural two-way AI counselling conversations * Structured treatment sessions and mandatory treatment rules * Adaptive questioning based on client responses * Long-term personalized client memory across sessions * Progress tracking, including cravings, confidence, nicotine use and treatment completion * Different pathways based on client progress, difficulties or lapse/relapse situations * Treatment exercises and supporting documents * Session scheduling, reminders and follow-up * Voice-based AI conversation * Multilingual operation * Web and future mobile-app integration * Administration and analytics Critical Requirement — Model Independence We do (not) want the treatment system permanently dependent on one AI provider. The architecture should allow us to use or migrate between: * OpenAI * Google Gemini * Anthropic Claude * DeepSeek * Future commercial or locally hosted AI models Our treatment methodology, treatment rules, client data, and business logic must remain separate from the underlying LLM. Changing the AI provider should not require rebuilding the treatment program. Treatment Content Must Be Separate From Code Our clinical team will continuously develop and improve the treatment. Authorized non-programmers should eventually be able to modify: * Treatment sessions * Questions * Instructions * Educational content * Treatment rules and thresholds * Supporting documents * Treatment sequences without requiring a programmer for ordinary content changes. Treatment version control will also be required. Long-Term Memory The AI should appropriately remember relevant information from earlier treatment sessions. For example, a later session may need to know: * Previous craving situations * Confidence scores * Difficulties the client experienced * Techniques previously used * Goals or commitments * Previous nicotine use or lapses * Relevant changes in the client's situation We are interested in a good architecture combining **structured client data, conversational/semantic memory, RAG/knowledge retrieval, and deterministic treatment rules**, rather than simply sending the complete conversation history to an LLM. Privacy, Security & Ownership Because the platform may process sensitive health-related information, privacy and security must be considered from the beginning. Experience or understanding of the following is important: * GDPR/privacy-oriented architecture * Secure authentication and access control * Encryption and database security * Backup and recovery * Appropriate handling of data sent to AI providers * Data portability and export (We must retain ownership and control of our treatment content, treatment rules and client data.) We also want to avoid unnecessary dependency on a particular developer, development company, cloud provider or AI provider. Good documentation and portability are therefore important requirements. Technical Experience We are particularly interested in experience with: * Generative AI and Large Language Models * AI agents and conversational AI * RAG / knowledge retrieval * Long-term AI memory * Vector databases / semantic search * Structured outputs and tool/function calling * LLM orchestration * API development * Software architecture * Python/backend development * Databases * Cloud infrastructure * Voice AI (STT/TTS) * Multilingual AI systems You do not need to be the leading expert in every area. More important is the ability to design the overall architecture correctly and explain your technical decisions clearly. Individual Developer or Small Team We welcome applications from either: 1. An exceptional senior AI architect/developer, or 2. A small specialized AI development team. If applying as a team or agency, please identify the (technical architect who would personally lead our project), explain their experience, and briefly describe the role of each proposed team member. We want direct access to the technical person responsible for the architecture — not only an account manager or salesperson. Initial Paid Test / Discovery Phase Although this may become a long-term engagement, we intend to begin with a **small paid technical/discovery assignment**. Shortlisted candidates will later receive a limited number of representative treatment sessions and client scenarios. We will use these to evaluate how your proposed architecture handles: * Natural conversation * Mandatory treatment instructions * Long-term memory * Structured client information * Adaptive treatment pathways * Content changes * Model/provider switching Detailed proprietary treatment material will only be shared at the appropriate stage. When Applying Please briefly answer these questions: 1. Describe one substantial conversational AI/LLM system you personally designed or built. What exactly was your responsibility? 2. How would you separate our treatment program from the underlying LLM so that we could change from OpenAI to Gemini, Claude, DeepSeek, or another model without rebuilding the treatment system? 3. How would you approach long-term client memory across treatment sessions lasting several months? 4. How would you allow our clinical team to modify sessions, questions, and treatment rules without requiring a programmer for every ordinary change? 5. If applying as an agency/team, who will personally be the technical architect responsible for our system? Please provide examples of relevant systems you have actually worked on and clearly explain (your personal role ). Please avoid generic proposals. We are much more interested in your architecture, reasoning, communication, and experience. Any generic proposals will be tested and rejected Long-Term Potential We intend to start carefully with a first working version and develop the platform progressively. If the collaboration is successful, this can become a substantial long-term project involving additional treatment functionality, voice AI, languages, analytics, mobile applications and international markets.