Consulting engagements

AI infrastructure and automation, built around the right model strategy.

I help teams design and ship AI systems using frontier model APIs, private or local LLM clusters, or hybrid architectures that combine both, based on the constraints that actually matter to the business.

Weighing API spend against your own GPUs? Try the GPU vs API cost calculator.

  • Review your goals, current environment, and delivery constraints at a high level.
  • Identify whether the next step should be a proposal, architecture assessment, or implementation engagement.
  • If deeper analysis is needed, scope the assessment before substantive architecture work begins.

Frontier model APIs

Use managed models when capability, multimodal support, and speed to value matter most. I help scope the provider, guardrails, routing, and operational shape around it.

Private and local clusters

Use self-hosted or VPC-isolated inference when ownership, compliance, air-gapped access, latency control, or cost at scale pushes the workload private.

Hybrid AI architectures

Route sensitive workloads to private models and general workloads to frontier APIs. Hybrid is often the practical answer when requirements conflict.

How engagements start

Free fit call first. Paid architecture work only after the scope is clear.

The initial consultation is for fit and scoping. Proposal conversations are part of the sales process. Substantive technical discovery and architecture work starts only after we agree on the paid scope.

01

Free initial consultation

A 30-minute conversation to understand the problem, constraints, fit, and likely next step. It is intentionally a scoping conversation, not a full architecture assessment.

02

Proposal and scope review

If there is a fit, we can review the proposal, assumptions, scope, and engagement shape without turning the sales conversation into a separate billable consulting session.

03

AI Automation Audit, then build

The paid first step is usually the one-week AI Automation Audit (from $1,500). Its fee is credited in full toward any build package booked within 30 days, so starting with the audit costs nothing extra if you go on to build.

Pricing

Clear starting prices, a credited first step, and an ongoing option after handoff.

Every package shows a “from” price. The final fee is confirmed in the written scope before any paid work begins, and it depends on integrations, environments, and data boundaries.

01

Start with the audit

After the free call, the recommended paid first step is the one-week AI Automation Audit, from $1,500. The audit fee is credited in full toward any build package booked within 30 days.

02

Build to a fixed scope

AI Ops Starter from $4,500, Agent Mesh from $8,000, and LLM Private Cloud from $14,000. Work that cuts across packages or needs phased delivery is scoped as Custom AI Architecture instead of being forced into a template.

03

Stay supported after handoff

Your team owns the system. If you want senior guidance as it evolves, the Advisory Retainer runs from $1,500/month, month-to-month, cancel anytime.

See it run

What a production agent team actually does all day.

Goals become owned tasks. Agents pull context, draft, and get checked by evals that can send work back. A person approves anything that matters before it ships, and every step lands in an audit log with its cost. This is the shape of the systems I design and hand over.

Goal

Clear today's support queue in under 5 minutes a ticket

Crew

Audit log

  1. 09:10:58handoff docked
  2. 09:10:58ops docked to recharge
  3. 09:10:58eval back on the gate
  4. 09:10:11ops sent the reply and closed T-101 in 3m 40s
  5. 09:09:24ops woke on heartbeat
Illustrative workflow. Agents, models, and costs vary by engagement; the structure (owners, budgets, eval gates, human approval, audit trail) does not.
  1. router picked up the goal: clear today's queue
  2. router split the goal into T-101, T-102, T-103 with one owner each
  3. retriever checked out T-101
  4. eval took its post at the review gate
  5. sentinel loaded the policy rules
  6. scribe woke on heartbeat
  7. retriever attached 4 refund-policy passages to T-101
  8. retriever brought the T-101 context to the scribe
  9. handoff rolled over to carry the draft
  10. handoff drafted a reply for T-101
  11. handoff submitted T-101 for review
  12. retriever checked out T-102
  13. eval failed T-101 on tone and sent it back
  14. sentinel: no customer data in the draft
  15. retriever found 2 similar past tickets for T-102
  16. handoff carried T-101 back to the scribe
  17. handoff softened the reply and cited the policy
  18. handoff resubmitted T-101
  19. router checked progress on T-102
  20. eval passed T-101: tone, policy and PII checks
  21. sentinel cleared the fix
  22. a person approved T-101 (refunds always need sign-off)
  23. ops woke on heartbeat
  24. ops sent the reply and closed T-101 in 3m 40s
  25. eval back on the gate
  26. ops docked to recharge
  27. handoff docked

Packages

Five ways to engage, depending on what you need to prove or ship.

These are paid outcome-oriented engagements, not hourly call bundles. Most teams start with the AI Automation Audit; its fee is credited in full toward any build package booked within 30 days.

Diagnostic & entry pointStart here1 week

AI Automation Audit

A deep-dive technical audit of your existing automation flows, LLM prompts, tool integrations, and data boundaries to identify immediate optimizations.

Price
From $1,500
Credited toward your build
The paid first step after the free call. The audit fee is credited in full toward any build package booked within 30 days.

Best for

Teams with existing automation or AI workflows that want to identify bottlenecks, security risks, and scaling opportunities.

  • Review of existing n8n/Make workflows, scripts, or agent prompts
  • Data leakage, credentials, and security posture inspection
  • Latency and cost-reduction opportunities report
  • Prioritized roadmap of high-value automation opportunities
Book the free initial consultation
Fastest path to value3-4 weeks

AI Ops Starter

A working AI automation stack built around n8n and an LLM backbone that fits your requirements, whether that means frontier APIs, private inference, or a hybrid route.

Price
From $4,500

Best for

Teams that want one real workflow in production with the right model layer behind it.

  • One business tool integration
  • One agentic workflow mapped to a real process
  • Deployment to cloud, self-managed infrastructure, or on-prem
  • Scoped data-handling recommendations for your environment
Book the free initial consultation
Advanced engagement4-6 weeks

Agent Mesh

A multi-agent system using MCP patterns so models can inspect and act on infrastructure safely, with explicit approvals for mutating operations and a clean audit trail.

Price
From $8,000

Best for

Platform and DevOps teams that want AI agents with scoped, auditable access to real infrastructure.

  • Scoped tool access for AWS, Azure, Kubernetes, Terraform, and internal APIs
  • Dry-run-first mutating workflows
  • Audit logging and approval gates
  • Model-agnostic routing across frontier, local, or hybrid stacks
Book the free initial consultation
Ownership first6-8 weeks

LLM Private Cloud

A full private deployment on client-owned or isolated infrastructure with model selection, inference endpoints, observability, and retrieval over internal knowledge sources.

Price
From $14,000

Best for

Teams that need to own the model runtime, data path, and deployment environment end to end.

  • Air-gapped or VPC-isolated deployment patterns
  • Inference endpoint and runtime configuration
  • RAG over internal knowledge sources
  • Observability, CI/CD integration, and documentation
Book the free initial consultation
Tailored engagementScoped after consultation or assessment

Custom AI Architecture

Custom-scoped engagements for frontier integrations, private infrastructure, hybrid routing, platform hardening, workflow automation, or existing AI systems that need production architecture.

Price
Custom

Best for

Teams with unclear architecture direction, migrations in flight, or requirements that cross multiple systems.

  • Scoping for mixed frontier and private workloads
  • Security, observability, and handoff planning
  • Support for existing Kubernetes, cloud, or internal platforms
  • A delivery shape aligned to your team and constraints
Book the free initial consultation
After handoffMonth-to-month, cancel anytime

Advisory Retainer

Ongoing architecture and model guidance once your team owns the system, with a regular review of how it is running and a fixed block of hands-on time when something needs building or fixing.

Price
From $1,500/month

Best for

Teams that own the system after handoff and want senior architecture guidance on call without hiring for it.

What is included

  • Ongoing architecture and model guidance as providers, models, and requirements change
  • Monthly review of cost, evals, and incidents
  • Priority async access, with a reply within 1 business day
  • Up to a fixed block of hands-on hours each month: a set number of hours, agreed up front

Ongoing guidance, not a managed service: your team keeps running the system. Month-to-month, cancel anytime.

Ask about the retainer on a free call

I work with scoped, revocable access and, where possible, in your own accounts.Security & data handling

At a glance

Package comparison table

A quick scan of deployment shape, timeline, and starting price.

PackageBest deployment shapeTimelinePrice
AI Automation AuditDiagnostic audit of existing automation flows, prompts, and APIs1 weekFrom $1,500
AI Ops StarterCloud, self-managed, or on-prem with frontier, private, or hybrid model routing3-4 weeksFrom $4,500
Agent MeshModel-agnostic agent layer across cloud or hybrid environments4-6 weeksFrom $8,000
LLM Private CloudAir-gapped or VPC-isolated inference and retrieval stack6-8 weeksFrom $14,000
Custom AI ArchitectureFrontier, private, hybrid, or existing client infrastructureScoped after consultation or assessmentCustom
Advisory RetainerOngoing guidance after handoff on whatever stack you now ownMonth-to-month, cancel anytimeFrom $1,500/month

Not a fit

This is not the right engagement for every team.

I would rather say no early than sell you the wrong shape of work.

You need full-time staff augmentation rather than a scoped consulting engagement.

You want a long-term operator running the system for you instead of a team handoff and internal ownership plan. (After handoff, the Advisory Retainer offers guidance, not operation.)

Your timeline is under three weeks and there is no room for discovery or production hardening.

Selected experience

Representative delivery work from prior professional experience.

These anonymized examples reflect prior engineering delivery. Claims are limited to what can be supported without exposing client-confidential details.

These examples come from prior professional work. New independent consulting engagements are presented separately and only with the client's approval.

On-Premise LLM for Financial Services

40% lower inference cost

Challenge

A financial services organization needed LLM capability without exposing proprietary data to commercial APIs under strict compliance constraints.

Approach

Designed and deployed an air-gapped LLM environment on client-owned infrastructure, including a custom retrieval layer over internal documents and a scoped tool-access pattern.

Outcome

Compared with commercial API usage, the deployment reduced inference cost by 40% while keeping full data residency compliance intact.

AWS Platform Consolidation for Multi-Tenant Operations

Lower per-site cost and less operational overhead

Challenge

Roughly 12 independently hosted WordPress properties were each running on separate EC2 instances, driving up cost and creating inconsistent security and maintenance practices.

Approach

Migrated the portfolio into a consolidated AWS ECS Fargate platform with Redis, EFS, ALB, CloudFront, WAF, Shield Pro, and New Relic observability.

Outcome

The result reduced per-site infrastructure cost and operational overhead while improving availability and security posture across all properties.

See how I build

A look at the implementation mindset behind the consulting.

Watch how I approach local inference, infrastructure setup, and developer-facing AI workflows in practice. The point is transparency, not theater.

Book the free initial consultation

Process

A scoped path from assessment to handoff.

After the free consultation, substantive technical work starts inside an agreed paid scope, usually the AI Automation Audit. Its fee is credited in full toward any build package booked within 30 days, and the build ends in a clean handoff to your team.

01

Audit or architecture assessment (1-2 weeks)

Usually the one-week AI Automation Audit, or a longer assessment when the architecture itself is undecided. It reviews your stack, data boundaries, security constraints, target state, and rough cost shape, and the output is a written recommendation and roadmap. The audit fee is credited in full toward any build package booked within 30 days.

02

Proof of concept (2-3 weeks)

A working deployment using your infrastructure, data, and integrations so you can evaluate the recommendation in a real environment before wider rollout.

03

Production handoff (1-3 weeks)

Runbooks, monitoring, CI/CD, and team transfer so the system can live with your engineers instead of staying dependent on the consultant. An optional Advisory Retainer is available afterwards for ongoing guidance.

Engineering principles

The consulting stance is simple: build what the requirement actually needs.

That includes saying when a frontier API is the better answer, when a private cluster is justified, and when hybrid routing is the cleanest compromise.

Right model, right boundary

The architecture follows the requirements, not my preference. Frontier when capability wins. Private when ownership wins. Hybrid when the real world needs both.

Auditability by default

Tool calls, pipeline actions, and delivery decisions should be inspectable by your team regardless of where the model runs.

Ownership at handoff

Code, config, workflows, and documentation belong to you. The goal is a usable system with a clean internal operator path.

No consultant lock-in

I do not optimize for dependency. I optimize for a production system your team can continue without me.

FAQ

Common questions before booking the call.

If you are still unsure whether this is a frontier, private, or hybrid problem, that uncertainty is normal and part of the early consulting value.

Can you build with OpenAI, Anthropic, Gemini, Azure OpenAI, or Bedrock?

Yes. The engagement can use frontier model APIs where they are the best fit. I help evaluate privacy, retention, vendor constraints, and routing so the provider choice matches the business and compliance context.

When should we use a private or local model instead?

When ownership, air-gapped access, compliance, latency control, or cost at scale dominate the tradeoff. Private inference is a requirement-driven choice, not a brand preference.

Can you design a hybrid setup that uses both?

Yes. Hybrid routing is often the best answer for teams with mixed workloads, where some requests belong on frontier models and others must stay private.

What if we do not know which architecture we need yet?

The free initial consultation is used to understand the problem and determine the right next step. If choosing the architecture requires deeper analysis, I scope a paid assessment before implementation; detailed comparisons, target-state design, diagrams, roadmaps, and estimates belong in that assessment rather than the free call.

Is the initial consultation free?

Yes. The first 30-minute consultation is free. Proposal review and scope clarification are not billed separately. Additional technical discovery, architecture workshops, advisory sessions, or working meetings are paid consulting unless they are already included in an accepted fixed-price engagement; scope is confirmed before that work begins.

What is the paid first step after the free call?

Usually the one-week AI Automation Audit, from $1,500. It produces a written assessment and prioritized roadmap, and the audit fee is credited in full toward any build package booked within 30 days.

What happens after handoff?

Your team owns and operates the system. If you want ongoing support, the Advisory Retainer (from $1,500/month) covers architecture and model guidance, a monthly review of cost, evals, and incidents, priority async access with a reply within 1 business day, and up to a fixed block of hands-on hours, a set number agreed up front. It is month-to-month and you can cancel anytime.

Do we need GPU hardware already?

Not unless private inference is definitely part of the engagement. For many teams, the fastest starting point is a frontier API while private requirements are validated during a scoped assessment or implementation engagement.

Can you work with our existing Kubernetes or cloud environment?

Yes. Existing infrastructure is usually the preferred starting point. If new platform work is needed, that gets scoped directly into the engagement.

Next step

Book the free initial consultation to choose the right engagement.

We will review goals and constraints, confirm fit, and decide whether the next step is a proposal, a fixed-price architecture assessment, or implementation. Detailed architecture work begins after the paid scope is agreed.