Control plane
AI and MCP Gateway
One governed endpoint for every model call and agent tool call, with routing, guardrails, token budgets, and audit.
AI transformation
AI Transformation
Silex designs, builds, and operates the infrastructure, automation, and governance that AI depends on. Our Forward Deployed Engineers work inside your environment, operate what they build, and transfer it to your team on your schedule.
The AI stack
Five layers, one team.
Each layer depends on the one below it. Silex designs and operates all five, so the engineers who build your GPU environment also govern the agents that run on it.
Forward Deployed Engineers: build, operate, transfer
Layer 5
Agents and applications
Agentic operations, retrieval-augmented generation, custom agents, and AI-enabled applications.
Layer 4
AI control plane
An AI gateway and an MCP gateway on Kong that govern all model and agent traffic, routed across self-hosted and hosted models.
Layer 3
Model serving
Open models served with vLLM and OpenShift AI, sized for your models, context lengths, and concurrency.
Layer 2
Platform and automation
OpenShift, Ansible Automation Platform, Terraform, GitOps, and policy as code.
Layer 1
AI infrastructure
GPU clusters, high-speed fabrics, storage, power and cooling, and the software that schedules and meters them.
Cyber resilience and observability
Read the stack from the bottom up. Model and agent traffic passes through the control plane on its way to your models and tools, and recovery and monitoring cover every layer. Forward Deployed Engineers work across all five layers.
Where to start
Seven offerings, one method.
Start with the layer that blocks you today. Each offering uses the same engagement model and connects to the others.
Operations
AIOps
Agents triage incidents and propose remediation, and people approve every change until a workflow earns promotion.
Platform
Automation and Platform Engineering
Ansible Automation Platform, Terraform, OpenShift, and GitOps, so every change runs through tested, approved code.
Infrastructure
AI Infrastructure
GPU environments designed, supplied, built, and commissioned, from facility requirements to signed acceptance records.
Model serving
Model Serving and Inference
Open models served on vLLM and OpenShift AI, with a sizing method that shows the calculation behind every GPU count.
Applications
Applied AI Engineering
Retrieval, evaluation harnesses, fine-tuning, and security testing that move a use case from proof of concept to a supported service.
Agents
Agent Engineering
Agents and agentic coding environments on Claude, Codex, and open-source harnesses, connected to your systems through MCP.
Human-in-the-loop by default
Autonomy is earned one workflow at a time.
Every agent in a Silex design starts at Assist. It investigates and drafts, and a person approves each change. A workflow moves up the ladder only when the agent's diagnoses match your engineers' resolutions and its failure tests pass.
Evaluate before you trust
Silex tests models and agents against realistic operations scenarios before they touch production, and continues to evaluate them while they run.
- Level 1
Assist
Agents investigate with read-only access and draft a diagnosis and a fix. A person makes every change.
- Level 2
Act with approval
A person approves each change in one step, and approved Ansible job templates execute and verify it.
- Level 3
Self-heal
Workflows that passed their failure tests run on their own, with every run recorded and a stop control.
How we engage
Build, operate, transfer.
The engineers who design your platform operate it, and then move ownership to your team.
- 01
Assess
A scoped workshop or assessment defines the use cases, current state, and success criteria.
- 02
Design
Silex produces an architecture blueprint with documented decisions and reviews it with your architecture and security teams.
- 03
Build
Silex deploys and configures the platform as code, tests it, and writes the runbooks.
- 04
Operate
Forward Deployed Engineers run the platform, onboard new use cases, and hold monthly service reviews and quarterly governance reviews.
- 05
Transfer
Silex moves operational ownership to your team through structured knowledge transfer, at the pace you choose.
Partners
The platforms we build AI on.
Silex engineers hold more than 160 vendor certifications. We recommend the platform that fits your requirements.
Certified Delivery Partner
Premier Partner
Select Tier Services Partner
Platinum Partner
Speaking
Silex CTO Derek Lewis speaks frequently at industry and customer events. At the State of Tennessee AI Summit, he delivered two keynotes.
Modernizing IT for GenAI at Scale
Secure By Design: Building Trustworthy GenAI Systems
Questions buyers ask
Questions
What does AI transformation mean at Silex?
It means putting AI into production under the controls your organization already requires. Silex designs, builds, and operates the infrastructure, automation, model serving, and governance that AI systems depend on, and then transfers operation to your team.
Do we need our own GPUs?
Not always. Many use cases run on hosted models through the Anthropic API, Amazon Bedrock, or Azure OpenAI. Self-hosted open models make sense when you want control over data, cost, and model choice. Silex sizes both options before you buy hardware.
Which models do you work with?
Silex works with Claude, GPT, Gemini, and open-weight models such as Llama, Mistral, and Gemma. The AI gateway lets your applications use any of them under one set of controls, so you can change models without rewriting applications.
How do you keep AI systems governed?
All model and agent traffic passes through an AI and MCP gateway that applies identity, guardrails, token budgets, redaction, and audit logging. Agents act only through approved automation, and a person approves each change until a workflow is promoted.
Can we start small?
Yes. Most engagements start with a workshop or a scoped assessment, such as an AI Governance Readiness Review or an Inference Sizing Assessment. From there, Silex takes one use case from pilot to production before you expand.
Next step
Book an AI workshop.
Silex-led workshops cover AI and MCP governance, agentic operations, model serving, and governed coding assistants. We scope each one to your environment.