Vector Database and AI Agent Specialist (Construction Sector Data)

About AI Metric
We’re a UK-based AI engineering agency that builds high-margin automation systems for the construction sector. Our mission is to turn decades of disorganised project data into searchable knowledge through AI and vector databases. Our clients include leading contractors navigating regulatory compliance, dispute defence, and operational planning.
Role Overview
We are hiring a skilled AI developer or data operations specialist to work across two key areas:
Project Vectorisation – Prepare SharePoint and Outlook data for storage in a vector database.
Demo Build – Construct a prototype chatbot that can reason across large document sets using semantic search.
You’ll work with both structured SOPs and open-ended architectural challenges. The role suits someone comfortable with Python, vector databases, and conversational AI workflows.
Key Responsibilities
A. SharePoint Vectorisation (Ongoing Work)
Extract and structure data from SharePoint, Outlook, and project archives
Clean and chunk documents into semantic units (~700 tokens)
Apply metadata to each document (project, source, type, date)
Output structured JSONL or Parquet files for vector embedding
Use enrichment tools to classify, tag, and QA documents
Support ingestion into a vector store for chatbot-based retrieval
B. Demo Challenge: AI Knowledge System Prototype (One-off Project)
We are exploring two solution paths for a large client with thousands of internal documents:
Option 1: Centralised Knowledge Chatbot
Aggregate documents into one searchable index using tools like LlamaIndex, Pinecone or LangChain
Build a conversational interface that retrieves answers across the entire dataset
Option 2: Distributed AI Agent Army
Divide documents into domain-specific clusters
Build multiple lightweight agents each specialising in a subset of knowledge
Allow queries to be routed to the most relevant agent
Your task is to build a lightweight demo using one of these options and record a short walkthrough explaining your approach, tool choices, and how it might scale.
Preferred Stack
GPT-4 or Claude for query reasoning
Python for document processing
LangChain / LlamaIndex / Weaviate / Pinecone (or your preference)
Streamlit or Gradio for the chatbot interface
OCR/parsing tools (PyMuPDF, Tika, Trafilatura, etc.)
Requirements
Strong Python development skills
Clear understanding of vector databases and semantic chunking
Familiar with SharePoint navigation and Outlook data structures
Experience structuring data for chatbot or Q&A models
Comfortable following detailed SOPs while solving open-ended technical challenges
Reliable, organised, and able to communicate in fluent written English
How to Apply
Please send us the following:
Your CV or OnlineJobs.ph profile
Indicate which demo approach (Option 1 or 2) you’d prefer to build, and why
Tools you’ve used for vectorisation, chunking, semantic search, or AI chatbot development
Your estimated timeline for demo delivery
Your hourly rate and weekly availability