AI Prototype to Reduce LLM Hallucinations

Project Overview

I am developing a lightweight proof-of-concept AI system focused on improving reasoning reliability and reducing hallucinations in generative AI workflows.

The core concept separates probabilistic AI generation from deterministic validation and structured knowledge verification. The goal is to demonstrate how generated outputs can be checked against governed constraints, relational logic, and validation layers before being accepted into a persistent knowledge structure.

This is NOT a request to build a full enterprise AI platform. The current objective is to create an MVP/prototype that demonstrates the architecture and validates core concepts.

Key prototype goals:

LLM-generated output workflows

Structured validation pipelines

Contradiction detection

Confidence filtering

Knowledge graph integration

Reasoning integrity checks

Simple dashboard or visual workflow demonstration

Preferred experience:

Python

OpenAI and/or Anthropic APIs

LangChain

Neo4j or graph databases

AI agents

RAG systems

Backend AI workflow architecture

I am looking for a practical engineer who can:

simplify architecture intelligently

recommend efficient MVP approaches

rapidly prototype concepts

communicate clearly and collaboratively

Initial budget is intentionally limited and focused on proof-of-concept development only. Potential exists for longer-term collaboration if the prototype demonstrates strong results.