Project Details
Owner
Humza Sami Chughtai
Release Date
2023 – Present
Services
AI Architecture, Full-Stack, DevOps
Duration
Ongoing
Overview
Weave AI is an enterprise AI platform I built solo — Azure, Kubernetes, React, and the full agent stack. Live across Finance, HR, Procurement, and Corporate.
- Reports, PowerPoints, charts, runtime analysis, and web search — not just chat
- Live Postgres and SQL Server retrieval via custom views built for LLM consumption
- Context layer on every agent so regulated answers stay grounded and auditable
- User-built micro-agents: pick warehouse tables or Qdrant sources (HR policies, contracts, etc.) within your access — compose and ship your own agent. Game changer.
Objective
Production AI in regulated environments — hallucination is not an option. I owned discovery, architecture, build, adoption, and ROI proof end to end.
- One person. Zero handoffs.
Process
Data & Agent Fabric
Structured data from finance, HR, procurement, and corporate systems flows into a central data warehouse; unstructured knowledge lives in Qdrant. Every source is permissioned by role — users only see what they can access. They select feeds and spin up micro-agents on top, each wrapped in a context layer for grounded outputs.
Infrastructure & Orchestration
Stood up dev, staging, and production on Azure — workflows promoted seamlessly across environments. n8n as the main orchestration layer: microservices composed inside workflows for high scalability, containers on Kubernetes to auto-scale as adoption and load grew.
Custom RAG Pipeline
Semantic, recursive, and structure-aware chunking — retrieval 70% above baseline, cutting hallucination risk in finance and HR.
Pre-built & Custom Agents
Shipped domain agents for PO matching, contract generation, policy Q&A, RFP analysis, and more — alongside self-serve builders for reports, decks, charts, and ad-hoc analysis on warehouse and vector data.
Adoption & Scale
30-40% platform adoption through hackathons, gamified rollouts, and an AI readiness programme that turned business users into champions.
Impact
Year-one results at enterprise scale.
- 60,000 hours saved annually
- 50+ production use cases across the business
- 70% lift in RAG accuracy vs. standard baselines
- 30-40% adoption — well above typical enterprise AI rollouts
- Multi-million AED in new revenue
- Delivered 5x faster than programmes split across consultants, architects, and developers