Isaac Attuah

I help AI startups get agents out of the demo and into production.

I'm Isaac Kofi Attuah, an AI/ML Customer Engineer on the Startups team at Google Cloud in Atlanta. Founders and engineering leaders bring me in for the hard middle of an AI build: how the agents fit together, which model earns its cost, and whether the capacity will be there on launch day.

Read my writing Get in touch
Isaac Attuah, smiling, wearing clear-framed glasses

Where I work in the stack

From what the user touches down to where the data lives.

  1. Agents

    Multi-agent architectures: who orchestrates, which steps stay deterministic, and how state is kept from leaking between tasks.

    Vertex AI Agent Builder, Gemini API

  2. Models

    Evaluation and selection. A representative eval on your own workload, then the smallest model that passes it.

    Model Garden, Gemini, Claude on Vertex AI

  3. Capacity

    GPU and TPU capacity and the inference infrastructure around it, planned before the traffic arrives.

    Provisioned Throughput, GKE, Cloud Run, Terraform

  4. Data

    The analytics and transactional platforms that agents read from and write to.

    BigQuery, Spanner, Looker

From the field back to product, and out to other engineers

What I learn with customers should not stay in one account.

4,600+Customer Engineers onboarded with the Google Cloud labs I built.
3,000+Cloud go-to-market staff using my certification guides.
550+attendees at the Agentic AI workshops I co-led at Atlanta Tech Week 2026.
360+engineers who studied from my Professional ML Engineer guide. I also led a cohort of 15 Customer Engineers through the exam.
$50K+cut from enterprise BigQuery costs, alongside ML and bug-tracking Looker dashboards shipped for large enterprise customers.

I also write customer-impact reports for leadership: what a Vertex AI tier or quota change does to real workloads, and where adoption is getting stuck.

Selected systems and civic work

Things I designed, led or built.

  • Customer Scenario Simulator II

    Designer and builder

    A practice tool for Customer Engineers. One agent plays a customer with a hidden objective you have to uncover through discovery. A second agent grades the conversation against a rubric, with evidence for every score, and BigQuery tracks how your skills move over time.

    Claude on Vertex AI, FastAPI, Cloud Run, BigQuery. Code on GitHub. How it works.

  • Internal Agentic Compliance Assistant

    Product and tech lead, team of two engineers

    A workflow agent that lives in Google Chat and keeps five regional pods on top of a recurring compliance process. Pods reached 98 to 100 percent compliance, the team got back roughly 900 hours a year, and the project was sponsored for an org-wide pilot.

    Gemini, Apps Script, Google Chat.

  • Google.org civic AI fellowships

    AI Coach, 2026. 20% Software Engineer, 2024

    In 2026 I coached in the Google.org Accelerator: AI for Government Innovation, advising Code for America and co-leading the design sprint at the D.C. kickoff. In 2024 I spent a fifth of my time building a generative AI business-licensing chatbot and web portal for the 3,000+ businesses of Macon-Bibb County, Georgia.

    Gemini, Google Cloud.

  • Terraform CI/CD Pipeline Kit

    Author

    A reusable Terraform blueprint that stands up a CI/CD pipeline deploying to both GKE and Cloud Run, so a new team starts with delivery already working.

    Terraform, Cloud Build, GKE, Cloud Run. Code on GitHub.

Writing

Field notes on agents, model fit and shipping.

All writing

Background

I joined Google in June 2022 as a Cloud Technical Resident and moved onto the Startups team as an AI/ML Customer Engineer in May 2023. Before that I studied computer science at KNUST in Ghana and then at the University of Miami. In 2026 I finished a master's in artificial intelligence while working full time.

The through line is teaching. Most of what I build ends up as a lab, a guide, a workshop or a practice tool for someone else.

Education

  • MSc. Artificial IntelligenceUniversity of the Cumberlands, 2026. GPA 4.0
  • BSc. Computer ScienceUniversity of Miami, 2022. Cum Laude, GPA 3.907. Prior study at KNUST

Building with agents, or stuck choosing a model? Write to me.

isaacattuah@gmail.com