Description
Summary We are seeking a Senior Full-Stack AI Engineer (or small team) to build a custom AI-powered personal knowledge and memory archiving platform on a strict 3-week timeline. The application enables users to upload personal writings and complete guided questionnaires, construct a grounded conversational AI persona, synthesize personalized voice and audio-driven avatar responses, and manage active versus sealed profile states. Project Timeline (3 Weeks Total) Week 1: Platform architecture, database schema, Next.js UI, guided Q&A engine, and document ingestion. Week 2: Grounded RAG integration (LLM APIs + Vector Store search), automated personality profiling, chat engine with confidence scoring, and user correction feedback system. Week 3: Voice cloning & TTS service integration, voice-driven interactive avatar lip-syncing, end-to-end testing, and deployment. Key Deliverables & Specifications 1. Dual-Mode Persona State Engine Active State: Editable mode allowing ongoing document ingestion, Q&A completion, chat testing, and output corrections. Sealed State: Read-only mode freezing persona parameters while maintaining full conversational fluency on novel queries. 2. Data Ingestion & RAG Memory Vault Guided Questionnaires: Categorized Q&A module with progress tracking. Document Vault: File upload pipeline (.txt/.md) with metadata tagging (era/category). Grounded RAG Pipeline: Vector Search indexing using LLM API vector stores with a local database fallback for grounded retrieval. 3. Automated Profiling Engine LLM pipeline analyzing uploaded content to extract structured personality traits, communication style guidelines, and decision heuristics. 4. Grounded Chat & User Feedback System First-person conversational interface powered by state-of-the-art LLMs. Calibrated Fidelity Rating: Real-time confidence indicator (High, Moderate, Low) based on retrieval grounding strength. Source Citations: Clickable UI anchors displaying source file titles and relevant text snippets. Human-in-the-Loop Corrections: In-chat panel enabling users to correct model responses and store feedback for model alignment. 5. Voice Cloning & Speech Microservice Reference audio upload with consent verification and safety safeguards. GPU-accelerated Python TTS microservice for low-latency voice cloning and multilingual speech synthesis. 6. Real-Time Interactive Avatar Sync Voice-driven 2D/3D avatar integration synchronized directly with synthesized TTS audio output. Core Tech Stack Frontend: Next.js, React, TypeScript, Tailwind CSS Database: PostgreSQL / SQLite AI & LLM: Google Gemini API / OpenAI API, Vector Databases & Embeddings Voice Service: Python, FastAPI, PyTorch, GPU-accelerated TTS Avatar Engine: Audio-driven lip-sync API or WebGL/3D animation pipeline How to Apply Please submit: Confirmation that you can commit to a strict 3-week delivery schedule. Portfolio or examples of past projects featuring Next.js, custom RAG pipelines, voice cloning, or interactive avatars. Proposed weekly milestone breakdown.