Lockmere AI Labs
The paperwork reads itself now.
We build the small tools that read your documents so nobody has to re-key them — for forwarders, brokers and importers of one to twenty people. Fixed price from $99, and every price is on the page.
- Client work
- Fixed price, from $99
- Timeline
- Days, not quarters
- Toolkit
- Vision, speech, classical ML
- Handover
- Yours to keep, no retainer
What we build
Two client practices, one product, one lab.
Tools and industry solutions are the client work, and they share an engine: the retrieval behind a tool is the retrieval behind an industry system. Companionship is our own product on its own track, and the lab is where we build the things we need and cannot buy.
AI Tools
Products built end to end — the retrieval layer, the model orchestration, and the interface a person actually touches.
- Retrieval that cites the passage it answered from
- Tool-using agents inside explicit permission scopes
- Evaluation suites that gate every release
- Model-agnostic — we route to whatever fits the task
Industry Solutions
Fixed-price custom AI for small forwarders, customs brokers, importers and distributors — commercial invoices, packing lists and bills of lading read, cross-checked and put where your team already looks.
- From $99, every price agreed before any work starts
- Delivered in days or weeks, handed over — no retainer
- Every field scored, anything doubtful flagged not guessed
- Measured on your own shipments before we hand it over
AI Companionship
Our own product line, on its own track: a desktop companion with emotional continuity — it remembers, it responds with warmth, and it is honest about what it is.
- Memory that carries across conversations
- Consistent character, not a reset every session
- A corner of your screen, not another tab to open
- Clear boundaries — never posing as a clinician
Research
VeilX, the messenger we built for ourselves: Matrix, end-to-end encrypted, hardened for the case where the threat is physical possession of the phone — and published as source you can read.
- Message store encrypted with Argon2id, keys in StrongBox
- Duress PIN that wipes instead of unlocking
- Locks on USB, SIM removal, or entering a geofence
- Server-signed remote lock and wipe
Since 2023, nearly every proposal in this industry has been a language model with a prompt attached. Most of these problems were solved better before that.
Our own product
A companion that remembers you.
The one thing here we are building for ourselves rather than for a client. A small character that lives in the corner of your screen instead of in a tab you have to remember to open — presence, not another place to go.
It carries context between conversations, because a companion that forgets you is a chatbot. Continuity is the whole product, and it is the hardest part to get right.
How we approach itIn development — a companion that occupies a corner of the screen, not your attention.
Research · in daily use
VeilX, an encrypted messenger anyone can inspect.
We built VeilX for ourselves rather than for a client — our own Matrix client and homeserver, because our own conversations should not sit with a vendor nobody can inspect. It is hardened for the threat that actually matters on a phone: someone else holding it, unlocked. The source is on GitHub.
A duress PIN that wipes instead of unlocking. Locking on USB, on SIM removal, on entering a geofence. Keys in a secure element and evicted the moment the screen goes dark.
What is in it- User ID
- @admin:lockmere.io
- Store
- Argon2id · StrongBox
- Homeserver
- tuwunel
Delegated from the apex domain, so identities stay short and the homeserver stays movable.
How we work
Three commitments we do not trade away.
Evaluated, not demoed
Every system ships with a measurable baseline and a regression suite. If we cannot show it got better, we do not claim it did.
A language model is one instrument
Half of what gets pitched as an AI problem is solved better, cheaper and far more reliably by vision, by classical ML, or by ordinary deterministic code. Knowing which is the engineering.
Honest about limits
A companion is not a therapist. A model is not an oracle. We say what our systems cannot do as clearly as what they can, including where they send your data.
Tell us what you are trying to build.
Send a short description of the problem and the constraints around it — data sensitivity, deadlines, what already exists. We reply with an honest read on scope and whether we are the right team.