Mark Marryatt

Helped pioneer Objective Foundation in 2023, building an AI harness before the patterns around agentic SaaS products had settled.

The platform grew into a multi-tenant SaaS application that combines reusable AI workflows, tenant-aware execution, controlled deployment paths and security boundaries suitable for public customer environments.

Focused on the platform layer that made this repeatable: infrastructure as code, tenant provisioning, CI/CD, Kubernetes deployment, identity, secrets, database access and operational guardrails.

Designed a repeatable Microsoft Fabric deployment model for a regulated data environment using Atmos and Terraform.

Provisioned and managed Azure/Fabric foundations including resource groups, Key Vault, Entra applications, Fabric capacities, domains, workspaces, lakehouses, SQL databases, Git integration, role assignments and shareable cloud connections.

Helped shape a governed lakehouse architecture with Bronze, Silver and Gold layers, source ingestion from systems such as M-Files, BambooHR and ClickUp, semantic models, Power BI reporting and Row-Level Security preserved through workspace separation.

Built the deployment approach around OIDC-authenticated GitHub Actions, environment-specific stacks, validation, drift detection and Fabric Git Integration so dev and prod data environments could be managed predictably.

Before I joined, the DevOps team struggled with slow and faulty deployments of a third-party, monolithic Java application onto virtual machine scale-sets.

Messy and imperative CloudInit scripts were used to unreliably stand up environments that required manual intervention to get to an operational state.

Redefined the process entirely, configuring CI to securely pull artifacts, build container images, test and publish helm charts and finally use terraform to stand up an instance of the application on Kubernetes in mere minutes.

Pioneered the usage of K6 as a load-testing tool, testing RESTful APIs as well as direct browser interactions. To quickly find resource sweet spots, the CICD pipeline created multiple ephemeral environments with different resource settings and tests would automatically be run against each.

Created highly informative dashboards from this data, linking commits, environments and resources to each test.

People gradually started following my methodologies in this area when they were able to see the value it provided.

Commissioned multiple systems to achieve a scalable automation solution.

Ansible AWX pulled playbooks and references to dynamic inventory sources from a Gitea repository, the dynamic inventory was enriched by device specific variables stored in Netbox.

The playbooks mostly automated router and firewall management. Configuration backup playbooks wrote device configuration to a dedicated Git repository, achieving version control and a level of insight not previously attainable.

Wrote these playbooks to be vendor-agnostic, using variables from Netbox to run the correct steps.

Other playbooks included the configuration of TACACS (Authentication/Authorization), standardizing ACLs and ensuring security practices are implemented.

Detailed dashboard were automatically created for each client, this information assisted greatly in reducing MTTR by making it possible to correlate events across devices, with information that would previously have been a painstaking ordeal to retrieve.

This implementation got me recognized as employee of the month - which was the third time during my time at Reflex.

This project saw a complete overhaul of the Elastic Stack in place, resulting in jumping multiple major releases from 2.3 to 7.10.

This included racking the physical equipment, creating virtual machines, installing the OSes, configuring RAID, cabling, networking, load balancing, security, logstash pipelines, index management, visualisations, dashboards and everything else in between.

Carefully planned the implementation of every node, automating a secure installation process.

Spinning up a new node and adding it to the cluster took only a few short minutes.

Different indexing and sharding strategies were tested by implementing Index Lifecycle Management, until the most effective combination was found.

This allowed for much more valuable alerts to be created, using in-depth KQL queries to output specific information to webhooks.

Logstash pipelines took the most development - enriching the raw data that came in to be much more useful.

The outcome was a secure and multi-tenanted platform thats employees and customers alike could access to view insights.

A FreeRADIUS implementation already existed at Reflex Solutions when I joined.

Took ownership of upgrading it to a new major version.

Containerized the application and ran it in production with Docker.

This project stemmed from the creation of an internal RADIUS frontend being built which led to me working closely in conjunction with our internal development team.

Throughout the process, we highly customized our RADIUS deployment. Created and troubleshot new SQL scripts to achieve our desired outcome.

Planned, scripted and executed the seamless decommissioning and replacement of core infrastructure in different regions.

These changes were applied flawlessly in minutes after weeks of meticulous planning and preparation.

Implemented AAA from the ground up, utilizing TACACS for Authentication and Authorization, and the Elastic Stack for Accounting.

All device logs were parsed using in-depth grok patterns and data enrichments, allowing for the creation of meaningful visualizations, providing holistic views into operational and security performance with the ability to drill down into specifics.

A large part of the redesign was separating core infrastructure into distinct management zones, enabling the application of fine-grained security policies to segments of the network at scale.