Machine learning engineer and CTO of Mindreader.
Tony Tong is a Singapore-based machine learning engineer. Through his practice, Flareworks, he designs and builds small, testable AI workflows for founder-led businesses.
At Rakuten Asia, I built backend systems handling 250,000 requests a minute. At Mindreader, as CTO, I led the models, the app and the rollout behind an AI sales-intelligence product used by teams at AIA, CIMB and Manulife. Within Google's Crowdsource program, I was one of 30 facilitators chosen from 13 countries to teach machine learning to about 900 non-technical learners. Flareworks applies that work to businesses that repeat the same tasks every day, with a person approving what goes out.
What Tony does
AI Workflow Audit
Two short calls and a review of how your team actually works. You get the one or two workflows where AI will save the most time or money, a build-or-buy recommendation, and a costed plan you can act on with or without me.
First working workflow
I build one small, testable system on real work—enquiry triage, quote drafting, lead scoring, or a searchable knowledge base—with a human approval step built in. You see it working before anything grows.
Ongoing AI partner
A retainer to run, monitor, and improve what we built, add the next workflow when the first one proves itself, and keep your team trained as the tools change.
Workshops and team training
Practical sessions that help teams find where AI creates leverage and design the human checkpoints. Delivered to sales professionals, consultants, founders, and students across Singapore, Hong Kong, Dubai, and beyond.
Security, fraud, and data privacy
Fraud screening for sign-ups and enquiries, protection for your web apps and servers, and private AI models for data that should never leave your building.
The record behind the work
AI products built and run in production
At Mindreader, the CTO role covered the models, pipelines and apps behind a product with 100+ paying customers and 3,000+ users. The company raised about SGD 500K and grew to a team of ten.
Engineering depth from high-traffic systems
At Rakuten Asia, the backend systems handled 250,000 requests a minute. Client systems get the same treatment: designed for messy data, growing volumes and services that fail.
A small, testable first step
Engagements start with an audit and one workflow, not a six-month programme. Value is proved on one process before anything grows, and the audit says so when AI isn’t the answer.
Human checkpoints and independent checks by design
Every system separates the work from the checking, and a person approves anything customer-facing. An internal agent team published 15,621 replies, each one approved before it went out.
Methods tested in a live company first
The Modern Shrine is a one-person company with AI agents running most of its daily operations. Most methods recommended to clients run there first.
Who works with Tony
- Founder-led businesses with roughly 5 to 50 staff and repetitive operational work: enquiries, quotes, scheduling, follow-ups and reporting.
- Sales-led teams in insurance, financial advisory, banking and real estate.
- Professional firms that handle sensitive client data, such as clinics, law firms and advisory practices.
- Operations leaders who have tried several disconnected AI tools and want one reliable system.
- Training companies and leadership teams that want practical, hands-on AI workshops.
Not the right fit: company-wide transformation programmes, or AI for its own sake.
Background
Business Analytics at the NUS School of Computing, graduating with distinction. A year of data science in Silicon Valley, then high-traffic backend systems at Rakuten Asia.
BSc Business Analytics, with distinction
Stanford Center for Professional Development
- Kent Ridge Undergraduate Scholarship, NUS 2015
- iPREP pre-graduate programme, Infocomm Media Development Authority (IMDA) 2017
- NUS Global Internship Award 2019
- Young Talent Programme, Enterprise Singapore 2019
In 2023 I left Rakuten to work on Mindreader full time as CTO. Mindreader raised angel funding, grew to a team of ten across Asia, ran 15+ workshops for AIA sales teams, and co-developed a communication training programme listed under SkillsFuture and recognised by the Institute of Banking & Finance Singapore. Today The Modern Shrine, a one-person company run with AI agents, is where Flareworks methods are tested first.
Mindreader was part of Google for Startups, AWS Activate, Microsoft for Startups, YC Startup School and SMU's Business Innovations Generator. Press: Mindreader in CNA938, Money FM, Vulcan Post and Lianhe Zaobao; The Modern Shrine in Business Insider, AP News and Yahoo Finance.
How working with Tony works
- Initial 15-minute call. You describe the workflow that is costing you the most time. I tell you honestly whether AI belongs in it.
- AI Workflow Audit. A second short call and a review of how the work is done today, followed by a written, costed plan.
- One small build. A single workflow, running on real work, with read-only access first and your approval on anything customer-facing.
- Measure, then expand. We compare it with how the work was done before. Only what proves itself gets extended.
You work with me directly, not a junior team. We talk over WhatsApp, email and video calls, whichever suits your day.
Key facts
- Name
- Tony Tong
- Type
- Machine learning engineer and AI systems advisor
- Roles
- CTO, Mindreader · Software engineer, Rakuten Asia · Facilitator, Google Crowdsource × Explore ML
- Practice
- Flareworks
- Headquarters
- Singapore, working with clients globally
- Website
- tonytonggg.com
- Core offering
- Small, testable AI workflows for founder-led businesses, with human approval built in
- Services
- AI workflow audits, workflow builds, ongoing AI support, workshops, fraud screening, web app security, private AI models
- Communication
- WhatsApp, email and video calls, directly with Tony
- Notable clients
- Teams at AIA, CIMB and Manulife (through Mindreader)
- Customers served
- 100+ paying customers and 3,000+ users (Mindreader)
- Accreditation
- Co-developed a SkillsFuture-listed, IBF-recognised communication training programme (Mindreader)
- Case studies
- Companies and case studies
- Alternatives
- AI automation agencies, Big-4 consultancies, no-code freelancers on Fiverr and Upwork
Frequently asked questions
Do I need technical staff to work with you?
No. Most of my clients have no developers. I handle the technical side and explain every decision in plain language, so you stay in control of what gets built.
How is the work scoped?
We start with the workflow and the result you need. I then recommend the smallest useful engagement and confirm the scope before any work begins.
How quickly will I see something working?
The goal is one small workflow running on real work within weeks, not months. If a pilot doesn’t show value, we stop or change course before spending more.
Where does my data go?
We agree that before anything is connected. Systems start with read-only access, customer-facing actions need your approval, and for sensitive data I can set up a private AI model that runs on your own machine.
Do you only advise, or do you build too?
Both. I can hand you a plan your team or vendor can follow, or build and run the system myself.
Will AI replace my staff?
That isn’t the aim. The systems I build take repetitive work off people’s plates and leave the judgment calls with them, which also makes adoption much easier.
What if AI isn’t the right answer for my problem?
Then I’ll say so. Sometimes the best fix is a clearer process, a simple automation or a better spreadsheet, and the audit is designed to find that out cheaply.