Conversational AI's Persistent Memory Layer

Build Persistent Memory Layer for Conversational AI (Python + DB + APIs)

We’re building a persistent memory system for a conversational AI project. The goal is to store and recall key text “memories” with both structured and semantic search.

What we need
• Backend service in Python (FastAPI or Flask).
• Storage: PostgreSQL (structured) + vector DB (Pinecone, Weaviate, Qdrant, or FAISS).
• API endpoints to save entries, recall by keyword or semantic match, and list recent.
• Dockerized deployment + clear documentation (setup guide + runbook).
• Secure handling of secrets (.env, vault).

Requirements
• Strong experience in Python backend engineering.
• Hands-on with vector databases + embeddings.
• Proven ability to deliver production-ready APIs.
• Comfortable working asynchronously and fully digitally (written updates, no Zoom).

Deliverables
• Working API + demo (3 saved entries, recall verified).
• <500ms recall latency, ≥90% recall accuracy.
• Clean repo with Dockerfile, README, and 14-day bug-fix warranty.

Budget & Timeline
• Budget: $2,000–$4,000
• Timeline: 7–10 days (stretch goal 5)