AI & Automation
Repetitive processing, document handling and routing move from manual steps to scheduled, monitored automation. We automate the parts of a process that are already well understood, then extend from there.
Learn moreHong Kong
AI, data and software solutions built for real-world operations.
We design and build AI-assisted workflows, data systems, automation and custom applications for the processes a business actually runs on. Reliability and long-term maintenance are part of the work, not a follow-up project.
- Focus
- Operational systems
- Delivery
- Design, build, maintain
- Approach
- Engineering-led
Capabilities
Four areas of work
Most engagements start in one of these areas and grow into the next one as the process becomes clearer.
Custom Software
Internal tools, dashboards, web applications and customer-facing systems, built around how a team actually works. Scope stays close to the operational problem being solved.
Learn moreData & API Integration
Databases, third-party APIs, spreadsheets and file drops connected into data flows that can be scheduled, retried and inspected when something goes wrong.
Learn moreAI Integration
Language models applied to specific tasks — classification, extraction, summarisation, drafting support — inside existing systems, with validation and fallback behaviour around them.
Learn moreHow we work
A short path from process to production
Four stages, each with a clear output. If a stage shows the project is not worth building, that is a useful result too.
- 01
Understand
We map the current process, its constraints, the data sources involved and what a good outcome looks like in operational terms.
Process walkthrough, data review, constraints, success criteria
- 02
Design
Technical architecture, workflow and integration approach are agreed before implementation, including how the system behaves when inputs are wrong or a dependency is down.
Architecture, data model, integration boundaries, failure handling
- 03
Build
Implementation happens in reviewable increments. Production concerns — validation, logging, access control, deployment — are part of the build rather than a later phase.
Incremental delivery, testing, deployment, documentation
- 04
Improve
Once a system is in use, behaviour and cost are measured against the original objective, and the next changes are prioritised from that evidence.
Monitoring, measurement, iteration, handover
Engineering
What we build with
No badges or vendor logos — just the areas we work in and the tools we use to do it. Choices are made per project, based on what the team can maintain.
Languages
TypeScript · Python · SQL
Interfaces
REST APIs · Webhooks · Server-side rendering
Data
Relational databases · Object storage · Structured logging
Platform
Cloudflare · Serverless runtimes · CI-based deployment
AI
Hosted AI APIs · Prompt versioning · Output validation
Architecture principles
- Static where possible
- Serverless where practical
- Clear API boundaries
- Least-privilege secrets
- Structured data
- Observable services
-
AI / LLM Integration
Task-specific model use with validation, fallbacks and cost visibility.
-
APIs
Versioned HTTP interfaces, webhooks and authenticated service boundaries.
-
Automation
Scheduled and event-driven jobs with retries, logging and exception queues.
-
Data Processing
Normalisation, validation and transformation across structured and text data.
-
Cloud Infrastructure
Serverless and edge deployment, environment separation, managed secrets.
-
Web Applications
Accessible, responsive interfaces with server-side validation throughout.
-
System Integration
Connections between operational systems with explicit ownership of each field.
-
Analytics
Operational reporting and measurement built on the same data the system uses.
Use cases
Where this work usually applies
Representative examples of the problems we are asked to solve. They are described as categories rather than case studies, because published client work requires client permission.
Workflow Automation
Manual, repeated steps become an automated workflow that can be monitored, re-run and maintained by the team that depends on it.
- Scheduled processing runs
- Approval and routing rules
- Exception queues
Internal Tools
Dashboards, operations consoles and management screens that match an existing process instead of forcing a team into a generic product.
- Operations dashboards
- Record management
- Role-based access
Data Processing
Structured and unstructured data from different sources is cleaned, transformed and brought together so it can be queried and reported on.
- Format normalisation
- Validation rules
- Scheduled exports
AI-assisted Operations
Classification, summarisation, information extraction and decision support added where the task is well defined and the output can be checked.
- Document extraction
- Triage and tagging
- Draft generation with review
API Integration
Payment, CRM, communication and data provider platforms connected with clear boundaries, error handling and retry behaviour.
- Payment and billing flows
- CRM synchronisation
- Notification delivery
Engagement
How projects usually start
Scope is agreed in writing before work begins. Pricing depends on the shape of the project rather than a fixed package.
Assessment
A short review of the process, data and constraints, ending with a recommendation and an estimate.
Defined project
Implementation of an agreed scope: automation, integration or an application.
Ongoing work
A predictable amount of engineering capacity for systems that keep evolving.
Discuss a project
Describe the process or system you have in mind. We will reply with an assessment of whether it is worth building, what it would involve, and what we would need from your side.