Built the shared control plane for AI agents, tools, knowledge bases and MCP servers, and orchestrated multi-agent workflows on Anthropic, and OpenAI models that execute customer SOPs, analyse documents, flag exceptions and suggest actions through chat and generative UI.

Helped re-platform to an event-driven microservices architecture that enabled company-wide adoption of agentic workflows, shipped to GCP on Kubernetes with Helm, Terraform, ArgoCD and GitLab CI at 80% test coverage, with observability via Grafana and Sentry.

Designed a config-driven ETL engine with a declarative UI that chains source components into custom inbound data flows, hydrating a single auditable form through a Kafka pipeline before pushing it to the customer's TMS; added agents to assist with config generation from prompts.

Built a quotes and rates platform implementing cost, surcharge and rebate logic from the ground up to generate any-to-any shipping quotes with full breakdowns for Toll Group and EmoTrans; cut peak memory from 8GB to 300MB (96%) and raised XML rate-processing throughput using asynchronous Celery workers and streamed parallel parsing.

Centralised shipment push across TMS integrations into a graph-based workflow builder that lets customers define their own workflows visually, backed by a template-driven response-parser service that maps unstructured third-party API responses onto client-defined schemas.

Built Python automation pipelines and scheduled jobs for system administration and server-health reporting.

Built an interactive user management portal for a CMS backend, hosted on Azure.

Developed the backend for a headless CMS platform, deployed as AWS Lambda serverless functions via GitHub Actions.