AI engineering studio · Hong Kong
Applied AI,built to runon your terms.
We design, build and operate AI systems for enterprises: from strategy and custom models to private on-premise deployments, and the protection of the intellectual property your AI creates.
Products
Software we build and own
Alongside client work, AIFLOW LABS builds its own products. Listed with their real status, no inflation.
RobotFlow Labs
Robot data and model development, built by AIFLOW LABS
A data and model platform for robotics: datasets, perception and control modules (ANIMA), and the pipelines to train, compress and ship them to real robots. Built entirely by AIFLOW LABS. It will become its own entity later.
- Robot datasets and perception modules (ANIMA)
- Training, distillation and export pipelines
- Live at robotflowlabs.com
FORGE
Provenance-enforced model distillation, compression and export
Turns large vision-language-action models into small ones that fit edge hardware. Provenance-tracked labels from registered teachers, multi-teacher distillation, control-aware pruning and quantisation, and validated exports to ONNX, TensorRT and Apple MLX, all through one CLI. Internal runs reached 28× compression; published numbers wait for the launch gates in the repo.
- 28× compression in internal benchmarks
- Multi-teacher distillation with provenance tracking
- ONNX · TensorRT · MLX exports, one CLI
Local AI Node
A private LLM appliance for SMEs
A pre-sized GPU box with open-weight models, retrieval over your own documents and a simple admin console. Plug it into the office network; nothing leaves it.
- Open-weight models, fine-tuned on your data
- On-premise, air-gapped option
- Fixed cost, no per-token bills
Provenance Toolkit
Watermarking and ownership proofs for AI models
Embeds verifiable signatures in model weights and outputs, fingerprints datasets and keeps an audit trail, so a company can prove what it owns if a model leaks.
- Weight and output watermarking
- Dataset fingerprinting
- Audit trail for ownership claims
Status labels are kept honest: “In development” means not yet shipped to a customer.
Services
What we do
Ten capabilities, one team. From the first workshop to a system running in production.
AI Advisory
Strategy, feasibility assessments and roadmaps for AI adoption that actually ships.
AI Consulting
Deep analysis of your processes and the design of the AI solutions that fit them.
Machine Learning Models
Custom models for prediction, classification and optimisation, trained on your data.
AI Agents
Chatbots, copilots and autonomous workflows that take action inside your systems.
AI-native Software
Products, APIs and platforms with AI at the core, engineered to scale.
Fractional CTO
Senior technical leadership for mid-sized projects that need direction, not just hands.
IT Consulting
Infrastructure, cloud and security advisory with an AI lens.
Process automation for SMEs
The work we are known for.
Small and mid-sized companies run on processes that were never designed, only inherited: quotes in spreadsheets, orders re-typed into three systems, paper on the factory floor. We automate them end to end, from the software your office uses to the hardware on your site, and the integration that turns them into one system.
Software
Workflows, document intelligence, connectors to ERP, CRM and accounting, approvals and AI agents that handle the routine work inside the tools you already use.
- Order-to-cash and quotations
- Invoice and document intake
- CRM, ERP and accounting sync
- Reporting and dashboards
Hardware
Scanners, sensors, cameras, label printers, PLCs and edge devices installed where the process physically happens: the warehouse, the shop floor, the retail counter.
- Barcode and RFID scanning
- IoT sensors and edge GPUs
- Vision inspection and counting
- PLC and machine interfaces
Integration
The part most vendors skip. We connect legacy systems, suppliers and devices through APIs and gateways so data is entered once and everyone sees the same truth.
- Legacy system bridges
- Supplier and logistics APIs
- Gateways and messaging
- Monitoring and alerts
Local AI
Your models. Your servers. Your data.
Cloud AI is fast to start and hard to control. We bring frontier-class open models inside your organisation: sized for your workload, fine-tuned on your knowledge, and served from hardware you own or rent privately.
Sovereign by design
Runs on-premise, in your VPC, or fully air-gapped. Built for Hong Kong PDPO and mainland PIPL data-residency requirements.
Fine-tuned on your knowledge
Retrieval and fine-tuning on your documents, tickets and data, so answers reflect your business, not the internet.
Predictable cost
A fixed-capacity GPU node replaces per-token invoices. We size it, deploy it and keep it running.
AI IP Protection
Own what your AI learns.
Models, datasets, prompts and agent workflows are now core company assets. We help you keep them yours, technically and contractually.
Model watermarking & fingerprinting
Embed verifiable signatures in weights and outputs so you can prove provenance if a model leaks.
Secure inference & deployment
Encrypted weights, hardened serving, access control and audit logs, from edge devices to data centres.
Licensing & data provenance
Clean training-data lineage, usage licences and contracts that hold up with partners, clients and investors.
IP strategy for AI
Which assets to protect, how, and where: patents, trade secrets, or open source with a moat.
We are engineers, not lawyers. We work alongside your legal counsel.
Approach
How we work
Two ways of working, chosen by what is at stake. A new system can be built the classic way. A running business cannot be stopped.
Discover
A two-week assessment: workshops, data audit and a ranked list of where AI pays off.
Prototype
A working proof on your real data in weeks, not quarters. You see it before you commit.
Deploy
Production rollout, integrated with your systems, with a security review before go-live.
Operate
Monitoring, retraining and support. The system keeps improving with your business.
Study
We sit with your users and map the process as it really happens, not as the manual says.
Build in parallel
The new flow is built beside your existing systems. ERP, accounting and databases are never touched.
Integrate with rollback
The two are connected through a controlled integration layer, with a rollback switch at every step.
Switch over
Users move to the new way at their own pace. No productivity stops, no big-bang migration.
Research & open source
More than consulting
We build our own technology and publish part of it in the open. It keeps our advice grounded in practice.
About
A Hong Kong AI studio with an engineer at the helm.
AIFLOW LABS was founded in Hong Kong in August 2025 by an engineer with more than 23 years in software. In its first 14 months it has served three active enterprise clients and generated about HK$1.2M in revenue. We stay small on purpose: the founder is in every project, from the first workshop to production.
Contact
Let’s talk about your project
Ready to put AI to work in your business? Tell us what you have in mind and we will get back to you quickly.