0
Producer services
PROJECT IN PROGRESS
Building a Data Mesh Platform with AI integrated
Research by Sushil Thesis Track
I am currently developing a platform where domain teams can register governed data products, stream events through Kafka, and make datasets discoverable through API and MCP tooling.
0
Kafka topics wired
0
Access layers (API, MCP, CLI)
0
Automated tests
THESIS PROFILE
Project context
Researcher
Sushil
Supervisor
To be announced
Thesis Window
Mar 2026 - Jul 2026
Current Stage
In active development
CURRENT IMPLEMENTATION
What I have already built right now
Contract Governance Core
A FastAPI contracts service with Postgres persistence, deduplication via unique indexes,
catalog search, product data preview endpoint, Prometheus metrics, and Kafka event emission
to data-contracts.
Multi-Domain Producers
Eight producers are available: MediaWiki, Jira, GitHub, Kimai, Paperless, Vtiger, Dovecot, and a secure generic producer with OIDC support.
MCP Integration
An MCP server and client stack enables tool-based discovery of contracts, schema reads, and data sampling, with Streamable HTTP support and resource URIs for domain context.
Observability + Delivery
Prometheus metrics across services, Grafana dashboard definitions, Docker Compose setup, Kubernetes manifests, and CI workflows for validation and image publishing.
SYSTEM SNAPSHOT
How the platform works
Domain Producers
→
Contracts API + Postgres
→
Kafka Topics + Consumer
→
MCP + CLI + External Agents
This is an active build, not a final release. I am iterating continuously on data quality, connector depth, and production-grade reliability.
ACTIVE DATA DOMAINS
Current producer landscape
📚 knowledge / mediawiki.documents
✅ work-management / jira.issues
📊 developer-observability / github.events
📅 finance-ops / kimai.timesheets
📄 records-management / paperless.uploads
👤 crm-ops / vtiger.contacts
💌 communications / dovecot.emails
🔧 generic producer / custom topics
THESIS ROADMAP
4-month execution timeline
Progress: calculating...
Window: --
Follow the build
I am building a practical blueprint for governed, discoverable, and AI-ready enterprise data products. This thesis demonstrates how contract-first thinking, streaming, and MCP interfaces can converge into a usable data mesh platform.