MarketPulse AI
AI agents & a RAG-based recommendation engine
System architecture
- 01
Real-time ingestion
Kafka + FastAPI — automated collection & processing
- 02
Autonomous agents
LangGraph orchestrates end-to-end workflows
- 03
Personalised RAG
LangChain + ChromaDB — answers per profile & context
- 04
Evaluated answer
DeepEval — multi-provider OpenAI / Mistral / HuggingFace
MarketPulse.AI combines a RAG-based recommendation engine (LangChain + ChromaDB) — personalising responses to the user profile and context — with autonomous AI agents (LangGraph) able to orchestrate complex end-to-end automated workflows. Data collection and processing are automated in real time (Kafka + FastAPI), and responses are evaluated with DeepEval on a multi-provider architecture (OpenAI, Mistral, HuggingFace).
What was built
- RAG-based recommendation engine (LangChain + ChromaDB) to personalise responses to the user profile and context
- Autonomous AI agents (LangGraph) able to orchestrate complex end-to-end automated workflows
- Automated real-time data collection and processing flows (Kafka + FastAPI)
- Response evaluation with DeepEval; multi-provider architecture (OpenAI, Mistral, HuggingFace)
The real challenge
The core challenge: making autonomous agents and a personalised RAG engine collaborate reliably, automating the whole real-time data flow — from collection to an evaluated answer.
Next project
Real Estate Fair Price Estimator