Making AI practical for organisations that deal in real information
Voxelith was founded on a simple observation: most organisations already hold the data they need — they just lack the infrastructure to interrogate it.
Back to HomeBuilt out of a genuine frustration
Voxelith came out of repeated conversations with finance and legal teams in Hong Kong who were sitting on years of accumulated documents, correspondence, and records — and had no practical way to query them. The tools existed. The willingness was there. What was missing was someone who could bridge the gap between AI capability and real organisational workflow.
We started in 2021 with a focus on knowledge graph construction for mid-size professional services firms. Over time, the work expanded to include text extraction pipelines for contract-heavy industries, and — as AI tools became more widely deployed — practical readiness training for teams who needed to work alongside these systems without becoming over-reliant on them.
Today, Voxelith operates from Wan Chai, working primarily with Hong Kong-based organisations but occasionally supporting regional engagements across Southeast Asia. Our team combines backgrounds in computational linguistics, information architecture, and domain consulting — which means we understand both the technical constraints and the operational context our clients operate within.
Our Mission
To reduce the distance between the information organisations hold and the decisions they need to make — through careful, domain-specific application of AI methods.
Our Approach
We scope narrowly, document thoroughly, and hand over work that your team can actually maintain. We're not interested in indefinite retainer arrangements or dependency — we'd rather you understand what you've been given.
Hong Kong Roots
Operating from Wan Chai puts us close to the financial district and the professional services firms that make up most of our client base. We understand the local regulatory landscape and the communication norms that affect how AI tools get adopted here.
A small team with a specific focus
We deliberately stay small — it keeps the work direct and the communication clear.
David Lam
Founder & Principal Consultant
Twelve years across computational linguistics and knowledge engineering. Previously led data architecture projects at a major Hong Kong law firm before founding Voxelith.
Sarah Chan
Lead NLP Engineer
Specialises in building text extraction pipelines for Cantonese and English mixed-language documents. MSc in Computer Science from HKUST.
Marcus Ng
Training Programme Lead
Designs and delivers the AI Readiness Training programme. Background in adult learning design and eight years in corporate training across financial services.
Standards we hold ourselves to
Data Confidentiality
NDA signed before any data is shared. We work under strict confidentiality protocols and do not retain client data beyond the engagement unless explicitly agreed.
Clear Scope Documentation
Every engagement begins with a written scope document. What's included, what's not, how success is measured, and what the handover looks like — agreed before work starts.
Documented Deliverables
All code, pipelines, and graph schemas are documented to a standard that allows your internal team to maintain and extend them without depending on us.
Honest Capability Assessment
We'll tell you when a proposed solution is unlikely to work or when a simpler approach would serve you better — even if it means a smaller engagement.
Privacy by Default
Compliant with Hong Kong's Personal Data (Privacy) Ordinance. Data minimisation and purpose limitation are applied throughout the project design phase.
Regular Communication
Progress updates at agreed milestones, not just at delivery. If something changes — scope, timeline, difficulty — you hear about it immediately, not at the end.
Why domain specificity matters in AI services
General-purpose AI tools are widely available. What's more difficult — and more valuable — is applying them with precision to the specific information structures and domain vocabularies that professional organisations work with. A contract management system at a law firm involves very different entity types, relationship structures, and extraction priorities than a research database at a life sciences company.
Voxelith's work is built around this specificity. We invest time in understanding how your organisation labels information, what relationships matter, and what the downstream use case actually requires. That investment shapes every design decision in a knowledge graph or extraction pipeline — and it's what separates useful AI tools from ones that produce technically correct but operationally useless outputs.
Our team's background spans computational linguistics, enterprise information architecture, and professional services consulting. That combination allows us to hold a technical conversation and an operational one at the same time — which is where most AI projects encounter friction.
Curious whether there's a fit?
A brief conversation is usually enough to tell. No obligation, no pitch deck — just a direct discussion about what you're working with.
Get in Touch