Hibiscus ConsultingThe short version
A CTO owns technical strategy. A fractional AI CTO does the same job on a part-time engagement, focused on the AI parts of your product, and unlike a pure advisor, stays hands-on enough to write and review the code that proves the strategy works. The distinction that matters for a small team: this is leadership plus individual-contributor depth. Direction that can't survive contact with a real codebase isn't worth paying for.
At Hibiscus that combination is the whole point. I led 17 engineers across 5 products at a neobank while writing core payments code directly. A fractional AI CTO engagement brings both to your team: the roadmap and architecture calls, and the pull requests that make them real.
What a fractional AI CTO actually does
The role is not "give AI advice in a monthly call." Here are the concrete responsibilities, the things that end up in your repo and your runbook:
Chooses the model-gateway architecture
Decides how requests route to models, where tool-using agents plug in, how you swap providers without a rewrite, and where cost and latency are controlled. One seam, not model calls scattered through your app.
Designs the RAG layer
Retrieval strategy, chunking, embeddings, and grounding, so the model answers from your data instead of hallucinating. Built to be measured, not just demoed.
Stands up evaluation harnesses
Backtests output against historical ground truth so you know when a prompt, model, or retrieval change actually made things better or quietly broke them. This is what turns a demo into something you can ship.
Defines safe agent tooling
Sets the rules for how agents call tools, touch data, and take actions, with the guardrails and permissions that keep a helpful agent from becoming a liability.
Sets the buy-vs-build and hiring calls
Which problems need a custom model, which need an API, what your first AI hires should look like, and what infrastructure is worth owning versus renting.
Ships the hard first slice
Writes and reviews the code for the piece that de-risks the rest, so the team has a real, deployed reference to build against, not a slide.
Auditable by default, not a black box
Most AI in production is a black box: something goes wrong and nobody can say why the model did what it did. Hibiscus builds the opposite posture. Every AI decision is signed, anchored, and independently verifiable, so you can reconstruct exactly what was retrieved, what the model saw, and what it produced, after the fact and without trusting my word for it.
The audit trail is real infrastructure: a tamper-evident log signed with post-quantum cryptography (ML-DSA, provisional patent filed). In regulated work, that turns "trust us" into "here is the proof."
If your AI touches money, health data, or anything a regulator or customer might question, this is the difference between a feature you can defend and one you have to apologize for.
Why leadership plus IC depth matters
Plenty of people can lead an engineering org. Plenty can write good code. The fractional AI CTO you actually want does both, because on a small team the person setting direction has to be the person who can prove it ships.
The leadership: three years running a neobank's engineering org across Stripe Connect multi-tenant, Plaid ACH, Apple and Google Pay, fraud rules, and KYC/KYB. The IC depth: I owned all commits on the Snap! Spend customer-facing React app and wrote core payments code across the platform. I also built a WebRTC platform for the Emmys in React in 5 weeks, and a React dashboard on a Rails EHR (Medaxion), so healthcare and roughly a decade of React/TypeScript are in the toolkit too.
What I build now is production AI end to end: model gateways routing tool-using agents, RAG, evaluation harnesses that backtest against historical ground truth, and real-time voice agents, including a live one you can actually talk to. Alongside a multi-tenant payments and field-ops platform (Stripe, Temporal, Postgres, Rust and Node on GCP).
When a fractional AI CTO fits (and when it doesn't)
Good fit
- Pre-Series A teams that need senior technical direction without a full-time executive salary
- Teams whose AI work has stalled at the demo stage and needs to become real, deployed, and measurable
- Products where AI touches regulated or high-stakes data and decisions have to be auditable
- Founders who can describe the problem but need someone to own the architecture and ship the first hard slice
Probably not yet
- You need a full-time executive in the room every day and can fund one
- You want advice only, with no one touching the code
- The problem is a well-scoped feature a contract engineer can just build (that's a different engagement, and I do those too)
- There's no real data or product yet to point AI at
If the good-fit column sounds like you, the next question is usually cost and structure. Those pages go deeper:
Frequently asked questions
What does a fractional AI CTO do day to day?+
They set your AI architecture and technical direction part-time while staying close enough to the code to ship it. Day to day that means choosing the model-gateway and RAG design, standing up evaluation harnesses that backtest output against historical ground truth, defining how agents use tools safely, making buy-vs-build and hiring calls, and personally writing or reviewing the code for the hardest first slice so the team has a real reference to build against.
When do I actually need one?+
Two common triggers. First, you're a pre-Series A team that needs senior technical judgment on AI but can't justify a full-time executive salary. Second, your AI work has stalled at the demo stage: it looks impressive but isn't deployed, measured, or trustworthy enough to put in front of real users. A fractional AI CTO gives you the direction of a senior leader and the throughput of a strong IC, sized to what you can fund.
How is a fractional AI CTO different from an AI consultant or advisor?+
An advisor gives opinions in a call and leaves the building to you. A fractional AI CTO owns the architecture and stays hands-on enough to prove it, writing and reviewing the code that ships. On a small team that hands-on depth is the point: direction that can't survive contact with a real codebase isn't worth paying for.
Why does "auditable by default" matter for AI?+
Because most production AI is a black box, so when something goes wrong nobody can explain why. Hibiscus builds AI where every decision is signed, anchored, and independently verifiable, on a tamper-evident audit trail signed with post-quantum cryptography (ML-DSA, provisional patent filed). If your AI touches money, health data, or anything a regulator or customer might question, that turns "trust us" into "here is the proof."
What kind of AI has Blake actually shipped?+
Production AI end to end: model gateways routing tool-using agents, RAG, evaluation harnesses that backtest against historical ground truth, and real-time voice agents (including a live one you can talk to). Outside AI, a Rust programming language (Nectar) that compiles to WebAssembly with a public compiler and 2,500+ tests, a multi-tenant payments and field-ops platform (Stripe, Temporal, Postgres, Rust and Node on GCP), and a post-quantum-signed audit trail. Career background: Director of Engineering at a neobank (~3 years, $720M in payments, 100% uptime, 17 engineers across 5 products).
Is Hibiscus remote, and can it do SBIR / federal work?+
Both. Hibiscus Consulting LLC works remote across the US and is Triangle-local in Raleigh/Cary, North Carolina. It's an SBIR-eligible small business, so federally-funded R&D engagements are on the table. Blake is also open to founding-engineer, senior/staff/director, and contract roles. Reach him at blake@hibiscus.buzz.