I'm Sam Keil — 15+ years turning ideas into measurable products. Founder, operator and hands-on builder. Now focused on helping senior teams build practical, ethical AI.
accreditationLovable Diamond (5-star) AI Developer
educationBachelor of Business — RMIT
ai stackClaude · Lovable · Supabase · Vercel · Cloudflare
About
I turn ideas into working, measurable products — then iterate with real data.
15+ years delivering pure digital and digitally-led physical products across fintech, retail, e-commerce and wellness. Comfortable moving between boardroom conversations, stakeholder workshops, Figma files, Supabase schemas and customer interviews in the same week.
I bring a rare blend: the strategic lens to shape an AI roadmap that fits a real business, plus the technical and delivery chops to actually build it — securely, cheaply, and with the right guardrails.
Right now I'm open to senior AI roles where that mix is useful — AI product leadership, head of AI, fractional AI leadership, or hands-on build projects. If that sounds like you, scroll down or reach out directly.
What I'm looking for
A senior AI role where I can ship.
Companies that take AI seriously but haven't drowned in the hype — looking for someone who can bridge strategy and build, and who cares as much about the risks as the upside.
Open to full-time, fractional, advisory, or fixed-term builds. Melbourne-based and open to relocating for the right role.
Work with me
Four ways to work together.
From full-time leadership to short, sharp builds — I can meet a business where it is in its AI journey.
01
Full-time senior AI role
Head of AI, AI Product Lead, Director of AI, or equivalent. Owning AI strategy and delivery for a business that's ready to put real resources behind it. My first preference.
● Available now
02
Fractional AI leadership
A day or two a week, for businesses not ready for a full-time hire but serious about doing AI well. Strategy, governance, roadmap, vendor selection, and hands-on where it helps.
● Taking one more engagement
03
AI consulting & advisory
Shorter, scoped engagements: AI opportunity assessments, governance design, responsible-AI playbooks, build-vs-buy reviews, or team coaching. Clear deliverables, fixed timeframes.
● Booking
04
Rapid build projects
I still love building. If you need a working AI prototype or MVP shipped in weeks not quarters — Lovable, Claude, Supabase, Vercel — I do that too. Best for validating an idea before committing serious spend.
● Limited capacity
Selected projects
15 years of shipping things that mattered.
PreCredits
2019 — PRESENT
Director · Fintech · B2B2C
End-to-end delivery of a regulated tokenised store-credit platform — from concept and regulatory model through product build, launch and iteration. Built with Claude + Lovable as the delivery method, Supabase-backed, Vercel-delivered. Three consecutive R&D Tax Incentive filings. Still running and iterating.
Stood up a boutique massage and wellness clinic end-to-end: site selection, lease negotiation, fit-out, brand, website, online booking, payments and day-to-day operations. A reminder that product thinking works just as well for a physical business as a digital one.
Founded Australia's first direct-to-consumer personalised pet food company, solo. Scaled to 30% month-on-month subscriber growth through 2018 — 93 → 240 active subscribers in just five months, with annualised forward revenue climbing from ~$65K to ~$170K in the same window. Led product, e-commerce, operations, marketing and funding. Sold shares and exited on favourable terms. Covered in SmartCompany and Inside FMCG.
Project Support Officer · Civic tech & stakeholder programs
Supported the City of Melbourne's Smart City Office — the team inside council responsible for the digital strategy, startup ecosystem and innovation programs that feed into the 10-year Future Melbourne plan. Coordinated a large stakeholder events program and contributed research on the city's startup ecosystem that informed council decision-making.
How I built Australia's first D2C personalised pet food company — solo.
The Petzyo story gets told in 2 sentences on most CVs. Here's the longer version — the decisions, the experiments, what worked, what didn't, and what the exit actually looked like. For the public-facing write-up, see the SmartCompany feature and petzyo.com.au.
Petzyo in the wild — brand shoot, warehouse fulfilment, raw-meal variants, sample-bag drop, and founder life mid-build (that's my sister's baby). Red brush-script PETZYO wordmark with a blush-pink + mint palette.
Peak growth
30% MoM in 2018
Role
Solo founder · concept → exit
Outcome
Exit under pressure · brand still trading
What transfers to AI product work
Personalisation is a pipeline, not a feature.
The Petzyo plan engine was a simpler version of what any AI recommendation or agent system has to do — signals, decisions, fallbacks when wrong.
Small experiments compound. Big bets rarely do.
30% MoM came from dozens of small tests, most of which failed. The same tempo I run AI pilots on now.
Own the orchestration, rent the infrastructure.
I didn't own a factory; I owned the brand and the operating spec. In AI: own the data access rules, prompts, and evaluation — not the model.
The quietest moat is how you handle being wrong.
Refunds, replacements, honest conversations — they built the brand. AI equivalent: how gracefully the system handles hallucination and escalation.
Ten chapters: concept, two business plans, prototypes, the investor memorandum, the hard exit — and the founder lessons most expensive to learn firsthand.
AI Philosophy
Fair, reciprocal, measured.
My approach to AI is built on being fair and reciprocal with the people who use it, acknowledging the intent behind every interaction, and staying cautious and measured with trust. AI systems earn trust incrementally — with transparency, observability, human oversight and the right guardrails before you turn the scale up.
"Most AI wins aren't about the model. They're about the boring, serious engineering around it — the data access rules, the prompt library, the fallback paths, the human review, the monitoring. Get that right and the rest gets easy."
How I build AI solutions
Simple architecture, strong guardrails, low cost to test — high ceiling to scale.
01
Data security via RLS
Row-level security in Supabase (or similar) so the AI can only see what the current user is allowed to see. Security isn't bolted on — it's a data-layer property.
02
Prompts & personality as documents
Agent style, tone and task prompts live in versioned, editable documents — not in code. Product and compliance can review and change behaviour without a deploy.
03
Model arbitrage
Use the right model for the job. Small, cheap models for classification and routing; premium models for reasoning and drafting. Easy to swap as the market shifts.
04
Cheap, fast, secure delivery
Supabase-style backends, Vercel-style delivery, Lovable/Claude for rapid iteration. Ship a safe prototype in days, validate with real users, scale only what earns its place.
What I bring
The skills I apply to senior AI work.
The capabilities I've built over 15+ years — and apply to every engagement, whether it's a full-time role, a short-term build, or an advisory brief.
Core
AI engineering, hands-on
Accredited 5-star (Diamond Level) Lovable AI Developer. Production software with Claude, Lovable, Supabase, Vercel and modern LLM tooling. Fluent from Figma through to deployed, monitored product.
Leadership
Stakeholder engagement & relationships
A career built on it — from founder negotiations at Petzyo and supplier deals, to coordinating external advisory teams at PreCredits, to delivering council-wide programs at the City of Melbourne's Smart City Office and senior-executive events at FACCI.
Cross-functional
Consulting across business functions
I've worn product, operations, finance, legal, marketing and customer support hats in the same week. I can sit with any team, ask the right curious questions, and translate their problem into something AI can actually help with — or tell them honestly when it can't.
Communication
Explaining complex ideas clearly
This very site is the evidence: a technical story, explained in plain English, with working demos. I've also written R&D Tax Incentive submissions for three consecutive years — a discipline in making complex technical work understandable to a non-technical audience.
Delivery
Shaping direction & seeing it through
Founded Petzyo solo, ran it to 20% MoM growth and exited on favourable terms. Took PreCredits from concept to operating fintech. Stood up Santal Organic Massage end-to-end — site selection, fit-out, brand, booking system, operations. I build what I plan.
Governance
Responsible & ethical AI
My default posture is cautious and measured. Data security via row-level security; prompts as versioned documents so compliance can review behaviour without a deploy; human-in-the-loop on anything affecting a customer outcome; model-agnostic design so risks can be isolated and controlled.
Collaboration
People-centred, collaboration & influencing
I start with the people using the thing — agents, engineers, customers — and work back from there. Influence by being useful: show a working prototype, invite critique, iterate. Trust compounds when the work is visible.
Commercial
Founder & operator instincts
I've signed the leases, raised the rounds, made the hires and the exits. Every AI decision gets filtered through a commercial lens — what does this cost, who pays, what's the opportunity cost, when does it break even? The boring questions that make the difference.
Interactive demo
AI opportunity explorer.
Pick an industry. See 2-3 AI opportunities I'd explore, with a realistic view of the business value and the risks that need controlling. All content is hand-written — no model calls, no API keys, no surprises.
Industry AI opportunity finder
A flavour of how I'd approach discovery with a new team.
LIVE
// All opportunities below are illustrative — a starting point for conversation, not a recommendation. Every real use-case would start with discovery, data access review, and a risk assessment.
Writing & thoughts
Short takes on building with AI.
Things I keep finding myself saying in meetings, turned into something I can link to. More coming.
DRAFT
RLS isn't AI infrastructure — it's AI safety infrastructure.
Row-level security is the cheapest guardrail in the modern AI stack, and the one people skip first. Why your data access layer is where AI safety actually lives.
~4 min read · coming soon
DRAFT
Your AI agent needs a prompt CMS, not a prompt file.
Prompts are product — they set tone, behaviour and risk posture. If compliance can't edit them without a deploy, you're doing it wrong. A small pattern, a big unlock.
~5 min read · coming soon
DRAFT
The cheapest AI strategy is the most honest one: ship less, watch harder.
Most AI projects fail not from lack of ambition but from too much of it. Why measured rollouts with real observability beat ambitious rollouts with hopes and dashboards.
~6 min read · coming soon
DRAFT
What 18 months of an AI-built fintech taught me.
Lovable + Claude + Supabase. What works, what's still hard, what I'd tell a founder starting today. Lessons from shipping PreCredits end-to-end.
