I am looking for an experienced LLM engineer (not just a UI developer) to build a custom AI persona application with: • stable personality + tone enforcement • long-term memory using embeddings or vector storage • manual model selection (non-negotiable) • persistent system-level instructions • session continuity across conversations This is not a simple chatbot wrapper. This is a full custom persona engine requiring precision, consistency, and deep API knowledge. ⸻ What I Need Built The final application (iOS/Web/Hybrid — open to your recommendation) must include: 1. Model Control Ability to select or lock the app to a specific OpenAI model version (e.g., GPT-5.1). The app must not default to Instant models. 2. System Prompt Architecture A persistent, hidden system layer controlling: • tone • narrative voice • emotional behavior • creative reasoning • persona consistency This can’t be editable by users. 3. Memory Module Long-term memory using: • vector embeddings OR • a structured memory file The AI must recall: • preferences • long-running projects • style rules • ongoing narrative constraints 4. Session Continuity Ability to save + resume conversations indefinitely. 5. Persona Enforcement Prevent drift toward generic AI tone. Use scaffolding, dual-layer prompts, or other techniques. 6. API Expertise You must use ChatCompletions API, not an Assistants wrapper. Must know how to handle: • temperature • top_p • penalties • prompt versioning • context windows ⸻ What I Will Provide • Full persona specification (“style codex”) • System-level instructions • Behavior rules • Tone/voice templates • Specific use cases • Long-term memory items Your task is to architect and implement the engine that ensures this persona is stable across sessions and model updates. ⸻ Required Skills Please only apply if you have experience in: • OpenAI API (advanced) • Vector databases (Pinecone, Chroma, Supabase, etc.) • Memory architectures • Prompt engineering • Persona modelling • iOS or web app development (React Native, Swift, etc.) • Session continuity • Building custom chat interfaces ⸻ What to Include in Your Application To filter out generic applicants, please answer these: 1. Which OpenAI models have you worked with directly? 2. Can you implement manual model selection in the UI? 3. Describe how you would enforce a stable persona over long conversations. 4. What memory approach would you recommend (embeddings vs custom)? 5. Have you built LLM personas before? Links welcome. 6. What’s your preferred tech stack for this? 7. What is your estimated timeline for an MVP? If you can provide a small technical explanation of: The difference between GPT-5.x and GPT-5.x-Instant in terms of persona stability …you move to the top of the list.