ProductEngineeringCloudAI Operations

Your fractional product & engineering department.

We partner directly with leadership to define, build, deploy, and operate complex software — delivering the capabilities of a mature engineering organization without the time and overhead required to assemble one.

Bring us the business problem See the delivery record
1,001merged pull requests
461production deploys
17repositories owned
828test & spec files
2×monthly output, May–Aug

One platform, April–August 2026. Measured from git and GitHub Actions history.

01What we do

One accountable team across the whole product surface.

Most teams can build. Fewer can build, ship, and then keep the thing alive at three in the morning. We do the whole arc, and verification is part of delivery — not a final-stage ceremony.

01

Full-stack delivery

Backend, web, and native mobile from one team.

  • NestJS microservices, APIs, payments and billing, routing, notifications, real-time systems
  • React customer experiences and operational dashboards
  • Native Swift/SwiftUI iOS and Kotlin/Compose Android
  • Typed API generation and third-party integrations
02

Quality engineered in

Tests that fail for real reasons, before your customers find them.

  • Unit, integration, regression, and production-matched container testing
  • Specialized harnesses for billing, Stripe, routing, migrations, and mobile behavior
  • Adversarial verification and traceable evidence, not green checks that prove nothing
  • Staging-first releases with production verification
03

CI/CD with guardrails

Automation that increases speed and control at once.

  • Fail-closed gates across linting, builds, tests, coverage, and migration safety
  • ECS preflight checks, controlled rollout, service-stability waits, build-marker validation
  • Audited emergency overrides and repeatable operational runbooks
04

Cloud reliability

Infrastructure designed around staying up.

  • AWS ECS, auto scaling, load balancing, ECR, RDS Proxy, Redis, Kafka, DynamoDB
  • Multi-AZ architecture, Terraform IaC, private networking, IAM, Secrets Manager
  • CloudWatch observability, actionable alerts, backup, recovery, incident response
05 · Compounding advantage

Institutional memory

The context does not leave when a person does.

  • An AI-assisted memory system retaining decisions, architecture, verification results, incidents, and operational context
  • Source-controlled and DynamoDB-backed, shared across people, agents, and sessions
06 · Tokenomics

Right model, right workload

AI spend governed by written policy, not habit.

  • Every agent workload declares its model tier before it ships
  • Cheapest adequate by default; reasoning-grade reserved for judgment and risk
  • Agent count and token cost recorded per piece of work

02Proof

Garbago: a platform, taken in-house and rebuilt.

A multi-tenant SaaS platform for waste haulers — customers, routing, crews, billing, and service delivery. It was built by an outsourced firm. We brought development in-house, rebaselined the roadmap against a code-verified audit rather than inherited documentation, and now run it.

What that took

A decommission and post-disengagement access review. A forensic billing audit across subscriptions, invoicing, commission, and dunning — separating measured revenue impact from presumed impact before spending engineering time on it.

A CI system built from nothing across eleven repositories, including a static probe that blocks a recurring defect class from corrupting money columns. A deployment pipeline with preflight checks, controlled rollout, and build-marker verification.

Then the production perimeter: load balancer locked to Cloudflare origin, bastion path isolated, and every credential exposed during the vendor transition rotated.

If you have inherited a codebase you did not write and cannot fully see, that is the exact problem we solve.

Services
7 NestJS microservices
Web
3 React applications
Mobile
Native iOS + Android
Infra
15 Terraform modules
Tests
828 test & spec files
Deploys
461 to production
Cadence
24/week average
Trend
177 → 334 PRs/month

03Hardware

We still make physical things.

ProtoShoff started in product design, prototyping, and light manufacturing, and that capability never left. It is why we can take a product from an enclosure to firmware to a cloud platform without handing you off three times.

Bridge project

Connected bin sensor

Where both halves of the practice meet.

  • ESP32 firmware — time-of-flight fill sensing, accelerometer, deep-sleep power budgeting, captive-portal provisioning
  • An LTE variant for deployments with no site WiFi
  • Printed enclosures iterated against real field mounting
  • Reporting into the platform that routes the trucks
Capability

Design and prototyping

Idea to a part you can hold.

  • Product design and CAD through to short production runs
  • FDM printing in PLA, ASA, ABS, PETG, nylon, carbon-fiber composites and polycarbonate
  • CNC milling in aluminium, titanium, hardened steels and alloys
  • Enclosure design for electronics that has to survive the field

04How engagements work

Bring us the business problem.

Not a spec. Not a ticket queue. The outcome you need and the constraints you are under — we will help define, build, deploy, operate, and continuously improve the technology behind it.

01

Define

Discovery with the people who own the outcome, then a written assessment of what exists, what is load-bearing, and what is quietly broken.

02

Build & ship

Incremental delivery behind fail-closed gates. Safe, small, ready any day — never a six-month reveal.

03

Operate

Monitoring, incident response, and runbooks, with the decisions and context recorded so the knowledge outlives the engagement.

Let's talk about what you are trying to build.

Long-term technology partnerships, fractional engineering leadership, and single-outcome engagements. Based in Gilbert, Arizona; working remotely with teams anywhere in the US.

Or reach us directly · mike@protoshoff.com · (480) 784-7770

Gilbert, Arizona Remote · United States linkedin.com/in/mikeshoffner