Tice LabsStart a project
AVAILABLE · Q3 2026 · 2 SLOTS

Ship the systems
that actually run the business.

Tice Labs is an independent consultancy turning manual operations into resilient AI-augmented workflows — backed by production AWS and Kubernetes done by one engineer who's been on call for it.

See client outcomes
10+
Years shipping infra
40+
K8s clusters
12
AWS regions
2
Max clients at once
~/projects/acme — tice cli
Production stack · battle-tested// 12 primitives
AWSKubernetesTerraformArgoCDOpenTelemetryBedrockLangGraphKarpenterDatadogSnowflakePostgresTemporal

// SERVICES

Three problems I solve.
All of them, end to end.

Most consultancies will sell you one of these and quietly subcontract the others. This is one engineer, working across the stack, with the scars to prove the seams are real.

01 / AWS

Cloud foundations that don’t surprise the invoice.

Greenfield landing zones or rescue work for accounts that grew faster than they were planned. Multi-account orgs, networking, IaC, and cost discipline.

  • Terraform / CDK from day one
  • Transit gateway + hybrid landing zones
  • Cost reviews with line-item ownership
EKS · Lambda · Bedrock · Step Fns
02 / Kubernetes

Production clusters, not classroom demos.

EKS, on-prem, and hybrid Kubernetes. GitOps, observability, autoscaling, and the boring operational work that determines whether the cluster runs you or you run the cluster.

  • ArgoCD / Flux GitOps pipelines
  • Karpenter, HPA, capacity planning
  • OpenTelemetry, Prometheus, runbooks
EKS · Istio · Cilium · Vault
03 / AI Workflows

Agents and pipelines that ship value, not novelty.

LLM systems wired into the work that already pays the bills. Evaluation harnesses, observability, and the unsexy reliability engineering that turns demos into dependable internal tools.

  • Agent + workflow orchestration
  • Evals, traces, regression suites
  • Cost and latency budgets enforced
Bedrock · LangGraph · Temporal

// COUNTER-POSITIONING

What I won't do.

The kinds of work I turn down — not because I can't, but because someone else would do them better. A short list of “no” is the most useful thing a consultant can publish.

  • Build chatbots

    If a chatbot is the answer, you don’t have an infrastructure problem. Hire a UX team.

  • RAG MVPs that die at the demo

    Most pilots aren’t worth productionizing. I’ll tell you up front, not at week six.

  • Train foundation models

    I wire production systems around models. If you need researchers, I know good ones.

  • Subcontract the work

    You hired one engineer. You get one engineer. The whole point of hiring me.

  • Fly-by audits as a deliverable

    I don’t show up, write a deck, and leave. The deliverable is shipped code and runbooks.

  • Open-ended retainers

    Engagements end on a date we agree to. The relationship can continue — the meter stops.

// OUTCOMES

Workflows, before and after. The numbers that get me hired again.

Support ops● SHIPPED

Tier-1 triage, automated

Manual
14m avg / ticket
Routed by humans · 2 swaps
AI-augmented
38s avg / ticket
Auto-classified · zero swaps
−96%Touch time
Finance ops● SHIPPED

Invoice extraction + GL coding

Manual
6 FTE · 4 days
Per close, with rework
Pipeline
< 90 min
Eval-gated, human-in-loop
38×Faster close
Cluster reliability● SHIPPED

EKS cost + reliability retune

Inherited
$48k/mo · 99.2%
Oversized, no Karpenter
Tuned
$19k/mo · 99.96%
Karpenter + PDBs + SLOs
−60%Spend, ↑ uptime

// HOW I WORK

The reasons clients keep coming back.

Tice Labs is one engineer with a decade of production scars. The operating model is shaped by what works at that scale — and what doesn't.

  • 01

    Direct access, always.

    No project managers, no tickets routed through three teams. You get the engineer doing the work — in your Slack, every day of the engagement.

  • 02

    Everything in code.

    Infrastructure, pipelines, runbooks, evals — committed to your repo, reviewable, reversible. You own a system at the end, not a dependency on me.

  • 03

    Built for handoff.

    The team after me should operate what I leave behind without a phone call. Docs, observability, and on-call posture are part of the deliverable.

  • 04

    Scoped, fixed, finished.

    Engagements have a defined outcome and an end date. If we need to rescope, we do it openly. No retainer creep, no parked invoices.

// ENGAGEMENT

How a project actually runs.

Typical engagement · 4–12 weeks · fixed fee
WEEK 0

Discovery & scope

A 60-minute call, then a written scope: outcomes, deliverables, fixed fee, end date. No-fit answers come back in 48 hours.

Statement of work
WEEK 1

Audit & baseline

Read-only access. I produce a written assessment with what’s working, what’s shaky, and the order to fix things. Numbers, not vibes.

Audit doc · baseline metrics
WEEK 2–3

Build the spine

IaC, pipelines, observability, eval harnesses — whatever the spine of the engagement is. Shipped to a staging environment you can poke at.

Repo + staging environment
WEEK 4–N

Ship to production

Cut-over plan, on-call posture, runbooks. We ship behind flags, with rollback. I stay on call until the SLOs hold for a full week.

Prod cut-over + SLO report
CLOSE

Handoff & teach

A live walkthrough recorded for the next engineer, plus written runbooks for the top 10 things that can go wrong. Then I leave.

Runbooks + handoff video

// THE ENGINEER

You're hiring me, not a logo.

JT
J. Tice · Founder
AWS CSA-Pro · CKA · 10+ yrs prod

I spent the last decade as a staff/principal engineer building the backend systems that other engineers depend on — multi-region AWS footprints, EKS at scale, the boring reliability work that keeps pagers quiet.

In the last two years that work shifted toward AI: agents answering tier-1 support, pipelines that turn unstructured intake into clean records, evaluation harnesses that keep all of it honest. The interesting part is that the engineering looks the same — observability, budgets, rollback. The hype is loud; the practice is familiar.

Tice Labs is what happens when I stop doing that inside one company and start doing it across a small number of clients per year, in scoped engagements I can fully own.

AWS Solutions Architect — ProfessionalCKA · Certified Kubernetes AdministratorEx-staff engineer (fintech / infra)

// FAQ

Questions worth answering up front.

How are engagements priced?
Fixed fee, tied to a written scope. We agree on outcomes and deliverables up front; the price doesn’t change unless the scope does, and rescopes happen openly. Typical engagements land between 4 and 12 weeks.
Do you work with my team or replace them?
I work with your team. I’m the embedded specialist for the duration — paired with whoever you have. The point is to leave the team smarter than I found it, not to create a dependency.
What about confidentiality and code access?
NDA-friendly, mutual or one-way. I work in your repos under your access controls. Anything I produce is committed to your codebase under your license — I keep nothing proprietary at the end.
Why “AI workflows” instead of just “AI consulting”?
Because most of the value isn’t in the model — it’s in the workflow around it. Evals, observability, rollback, cost ceilings, the human-in-the-loop edges. The model is a component. The system is the product.
Where are you based, and what hours?
Remote, US time zones. Default working overlap is 9–4 PT with flexibility for incident response. I take on at most two clients at a time so you actually get the engineer you hired.
What if we just need a second opinion?
Sometimes the right deliverable is a written audit and an opinion, not a build. Those are available as a fixed two-week engagement.

// CONTACT

Have a problem worth solving?

Send a paragraph or two on what you're building, what's broken, or what's stuck. I'll reply within two business days with whether it's a fit — and if it isn't, who I'd point you to instead.

hello@ticelabs.dev
Engagements 4–12 weeksRemote · US zonesNDA-friendlyReply < 48h