AI product innovation leader, with founder and commercial experience
I take novel ideas from first 
to paying customers.
For my whole career I've worked where technology, product, creativity and marketing meet. Since 2023, AI has been my obsession. I help teams find the ideas worth building, turn them into products people can test, and see them through to real use.
Open to senior AI innovation and new-venture roles: hybrid in Melbourne or Sydney, or remote in Australia.
- Venture-backed founderRead in a record. 500 Startups Seed funding from two 500 Startups funds, 500 TukTuks and 500 Durians, plus angel investors.
- Applied AIFrom memory. Since 2023 Foodkit's AI lab in 2023, then DeepBliss, a live AI product.
- Product builderFrom memory. Since 2005 From TD Waterhouse's broker interface to Sensis's ClickManager, then my own companies.
Brands I've built for in my own ventures
2010 to 2024
- Illustration: Nick sketching a product flow at a drafting table
- Illustration: Nick laughing with two colleagues at a work table
- Illustration: Nick walking a team through an AI workflow on a whiteboard
The Junction · 2010 to 2011
Nando's
Suncorp
Interflora
Collingwood Football Club
Aptivate.me · 2011 to 2015
Universal Pictures
Movember
Just Car Insurance
Mediacorp
Suncorp Bank
Bingle
Vibe Hotels
Foodkit · 2016 to 2024
Domino's- Nando's
Burger King
Swensen's
Minor Food
The Coffee Club
Taster
Soho Pizza
S&PAI lab
My selected work: My three most recent businesses
My latest project: DeepBliss
Founder & builder
From the deepbliss.me homepage
I designed, built and run DeepBliss with a team of AI agents.
A personal practice app. You describe a goal, and DeepBliss turns it into affirmations spoken in a clone of your own voice, layered over entrainment audio, with images and video of you living the future you described.
Live. Paying customers since July 2026.Visit deepbliss.me (opens in a new tab)
Vision board 
The Oracle 
Your sound, layered
Demos from deepbliss.me.
- The bet
- That a practice built around your own goal, in your own voice, would mean more to people than a generic recording.
- My role
- The whole product team: product and brand, the Oracle that turns an intention into a script, the voice and audio pipeline, image and video generation, billing, support and marketing.
- How I built it
- With AI agents under written rules. Claude Code writes the code, OpenAI's Codex reviews it, automated checks must pass before anything merges, and I decide what merges and what ships.
- Where it is now
- Live, with paying customers since July 2026. 278 merged changes since May 2026 and 54 production releases since July (counted October 31, 2026).
- What I learned
- An agent saying "done" isn't evidence. So I built the checks that make "done" mean something: independent review, automated tests, and a release step only I can run.
Three stories from DeepBliss
-
01
One control for music
What it showsProduct judgment from real use, and changing course when the evidence says the design is the problem.
- Shipped
- Tested
-
02
Five days, brief to release
What it showsHow I direct AI builders: they build and check, and the decisions and the release stay with me.
- Shipped
- Tested and reviewed
-
03
Billing in one list
What it showsCommercial ownership of the unglamorous details customers judge you on.
- Shipped
- Tested
[ Product film ]
Programmatic video created with AI.
DeepBliss explainer · 1:35 · May 2026
A short film from the DeepBliss homepage: one selfie, three visual styles, and the same AI-generated persona as the star of each.
My previous business: Foodkit
Co-founder & CEO
I built the product. I ran the company.
Foodkit gave restaurant chains their own branded ordering: web and mobile apps, in-store operations, call-center tools, loyalty, and an open API with 50+ integrations into delivery fleets, payments and point-of-sale.
"Foodkit provides the platform flexibility, stability and security our business needs to operate at the highest possible level."
[ The platform ]
A complete ecosystem of tech.
We gave big-name food and beverage brands fully white-label apps and websites, in-store kiosks and vendor consoles, social ordering, a developer API and a sophisticated set of modules.
Ordering channels
- Web
- Mobile
- Kiosk
- Social
- Voice
- Other
Foodkit headless commerce APIs
Platform modules
- Order management
- Outlet management
- Delivery management
- Marketing tools
- Designer tools
- Shopping cart
- Menu manager
- Online payments
- Content management
- Billing system
Add-on modules
- Loyalty & rewards
- Order manager Pro
- Customer service Pro
- Delivery driver app
- Kitchen display (KDS)
Integrations and partners, a selection
- Slack
- Intercom
- AWS
- Stripe
- Adyen
- GetSwift
- Google Maps
- Revel
- Twilio
- Infobip
- OneSignal
- Google Analytics
- Lalamove
- Mailgun
- Braze
- SevenRooms
- 2C2P
- Mixpanel
- Deliveree
- Omise
Foodkit data lake
[ The company ]
From seed round to 40+ people.
I raised seed funding from two 500 Startups funds, 500 TukTuks and 500 Durians, and from angel investors, set product direction, won the enterprise accounts, hired and led a team of more than forty at its peak, and took the platform into new markets.
500 TukTuks, e27, 2016500 Durians portfolio, archived 2019
What it showsBuilding a company, selling to enterprises, and leading teams across engineering, product, sales and operations.
- 8.35Morders
- US$320M+order value
- 40+people at peak
- 8markets
- 50+integrations
Delivery technology across markets
UK / LondonBrusselsParisQatarSouth AfricaHong KongThailandPhilippines · operating market
Select a market to highlight its location.
01From memory.
Eight markets, three continents.
Foodkit began as Ginja, a delivery business in Bangkok. Delivery was the hard part: we worked across third-party fleets, used incentives and game mechanics to keep drivers taking our jobs, and cut average delivery times 17% with routing and process work. What we learned became Foodkit's delivery tools. Restaurant chains went on to run Foodkit in eight markets across Europe, Africa and Asia.
What it showsTaking a product from one city to customers on three continents.
-
02From memory.
The AI lab, 2023
In 2023 I started Foodkit's AI lab: two of us, me included. With the call centers of Minor Food Group and S&P, we built and tested chat ordering, voice ordering and customer-service tools on Claude 2 and voice synthesis, and judged the models of the day not reliable enough in Thai.
What it showsKnowing when not to ship.
How it ended
Foodkit closed in 2024 after eight years. At the start of Covid my co-founder and CTO became seriously ill and spent a long time in and out of hospital. The platform's architecture and its enterprise integrations lived largely with him, and the business never fully recovered from his absence. What I'd do differently: write the system down from day one, so no company depends on what one person carries in their head. That lesson is why DeepBliss runs on written rules and decision records.
The one before that: Aptivate
Founder
Making brands playable.
I built interactive Facebook games and competitions that invited audiences into a brand's experience. Aptivate made competition and social apps for well-known brands, including Universal Pictures, Movember, Just Car Insurance, Suncorp Bank and MediaCorp in Singapore.
Just Car InsuranceRead in a record.
Rate My Ride
Three editions connecting car photos, community votes and leaderboards. Historical campaign reporting: 2.87M unique views, 38,000+ votes, 2,500+ entrants and 7+ minutes average engagement.
Universal Pictures
American Pie: Reunion
A character-matching experience: participants nominated friends they associated with the film's characters, using the Facebook sharing tools of the time.
Movember
Generation Moustache
Campaign creative and sign-up experience.
Bingle
Win a Jeep Cherokee
A promotional competition with trivia and a leaderboard.
Vibe Hotels
Get Vibe-rant
A hotel campaign with prize selection and entry forms.
Three companies taught me one thing.
[ Thesis ]
Building is no longer the hard part. Judgment is.
Agents can now write, test and ship a product faster than a founder can tell whether it should exist. So the scarce skill is judgment: choosing the right problem, testing the answer with real people, and knowing when to stop. That's the work I lead.
So how do I decide what's worth building?
[ Method ]
How I turn uncertainty into useful products
Six stages, with research, documentation and evaluation running through all of them. Work can start with a new idea or with a product that's already live, go back a stage, or stop. The more uncertain the bet and the costlier a mistake, the more rigor each stage gets.
{{s.num}}{{s.name}}{{s.q}}
- {{r.k}}
- {{r.v}}
- Example
- {{s.ex}} {{s.exSr}}
Depth scales with uncertainty and the cost of being wrong. The method works in a small team or across an organization, with each stage owned by the right people.
The AI layer, for the technically curious
In DeepBliss today
Independent review
Claude Code writes the code and OpenAI's Codex reviews the pull requests. Since mid-September 2026, Codex has reviewed 228 of the 229 changes I merged and raised at least one blocking or important finding on 135 of them. Agreement between models isn't proof. Tests and real use decide whether something works.
In DeepBliss today
Skills
A skill is a written procedure an agent follows for a recurring job, so I don't explain it twice. Mine cover starting a bounded task, mapping a change to the tests it needs, turning my test notes into tracked fixes, handing over between sessions, and releasing. Every one of the 24 test batches I sent between August and October 2026 was reviewed, 19 of them within an hour.
In DeepBliss today
Bounded loops
Agent work loops through build, test, review and fix, and stops at fixed points where I decide: before anything merges or ships, and before spending, customer data or anything irreversible. Automated loops have hard limits: a stalled audio job is retried twice and then refunded, and a release approval expires after three hours.
In DeepBliss today
Orchestration
One coordinating Claude session plans and reviews, separate build sessions each work in their own copy of the code, and a separate session prepares releases. Work in progress lives on shared boards and in dated handover notes. I haven't measured whether this makes the work faster.
In DeepBliss today
Project context
The agents work from 15 written rule files, 28 decision records and 60 skills, and the rules point back to the reference documents and decisions they come from, so each task starts from the context it needs.
In DeepBliss today
Learning
Failures become rules, tests or better procedures. When three of my agent's safety tests turned out to pass without checking anything, we added a rule: every new guard is broken on purpose once, to prove it can fail.
What I'd bring to a team
The same habits work with people in the seats: clear rules, visible evidence, fast loops, and the senior decisions kept with someone accountable for them.
Tool details as of October 31, 2026.
I learned it the long way.
The journey so far
Three chapters, each with something I still use. Every skill joins the line where I picked it up and runs to today.
2001 to 2010
Learning what moves people
What it gave me: how audiences behave, and how to test what works.
- 2001 · Kiss FM · Co-head of production, as a volunteer, while the station applied for its license. From memory.
- 2002 to 2008 · RRR · Volunteer announcer: almost six years on air, with music and interviews with international DJs. From memory.
- 2005 to 2007 · TD Waterhouse · Investment coordinator. I helped redesign the broker interface. From memory.
- 2007 to 2008 · Sensis · Product development manager, search. ClickManager had 1,400 customers in its first year. From memory.
- 2008 to 2009 · STA Travel · Senior digital marketing manager, APAC. My team of four was named Global Digital Team of the Year 2008. From memory.
- 2009 to 2010 · Tourism NT · Head of digital marketing, leading a team of 13. Generated over $420 million in attributable tourism expenditure for the Northern Territory through the Mix It Up Darwin campaign, according to Tourism NT's internal campaign reporting. From memory.
2010 to 2024
Building companies
What it gave me: how to build a company, sell to enterprises and lead a team.
- 2010 to 2011 · The Junction · Co-founder of a Melbourne agency for Nando's, Suncorp, Interflora and Collingwood FC. From memory.
- 2011 to 2015 · Aptivate.me · Founder. Facebook campaigns for Just Car, Suncorp Bank, Universal Pictures and MediaCorp. Read in a record.
- 2015 · TechGrind · Entrepreneur in residence, Bangkok. From memory.
- 2015 to 2016 · Ginja · Co-founder and CEO. Bangkok food delivery, backed by 500 TukTuks in 2016. Read in a record.
- 2016 to 2024 · Foodkit · Co-founder and CEO. Customers included Domino's and Nando's. Closed in 2024. Read in a record.
2023 to now
Building with AI
What it gave me: how to build with AI, and when not to.
- 2023 · Foodkit AI lab · Tested AI ordering with Thai call centers and judged the models not ready. From memory.
- 2024 to now · DeepBliss · A live AI product I designed, built and run with AI agents. Read in a record.
Next:
with your team.
Audience and story
Product and testing
Company and commercial
Applied AI
Solid: measured.
Outline: read in a record.
Dotted: from memory.
What I bring to a team
01
Find the opportunity
I saw that Yellow Pages sales teams could sell paid search: ClickManager had 1,400 customers in its first year.
02
Imagine differently
Affirmations in a clone of your own voice, with your future self on screen.
03
Make the complex clear
When an audio shortcut misfired in real use, I gave music its own control.
04
Build the first version
DeepBliss, end to end: product, voice pipeline, image and video, billing and support.
05
Orchestrate and challenge
One agent builds, another reviews, the tests decide, and I release.
06
Turn it into a business
Seed funding from two 500 Startups funds, enterprise customers including Domino's and Nando's, eight markets.
How I like to work
I'd rather show a prototype than a deck. I write the rules down so a team can move without me in every loop. I'm happy to be led, and when I lead, I ask the team what they think before I decide. When the evidence says I'm wrong, I change course.
What I bring from making things
Creative judgment with a producer's hands. I've made games, campaigns, radio and music, and now generative audio and video, so I can tell when a generated thing is finished and when it only looks finished.
