AlgoMaschine Intelligence

A project team for private legal AI · Based in Dubai

Private AI for legal work, built around your lawyers and kept within your own walls.

Write to us or book a call

Start with a conversation

One of us replies personally within one working day.

Book 20 min

Example of a contract review Illustration
12.3

Limitation of liability

The Supplier’s total liability under this Agreement shall not exceed the fees paid in the preceding three (3) months.

  • FoundThe limit is three months of fees. Your company’s rules ask for at least twelve.
  • SourcePage 14, clause 12.3
  • RuleYour contract rules, rule 4.2
  • SuggestsChange the limit to twelve (12) months.
  • DecidesYour lawyer. No change is made without approval.
18.1

Governing lawالقانون الواجب التطبيق

This Agreement is governed by the laws of the United Arab Emirates as applied in the Emirate of Dubai.

يخضع هذا العقد لقوانين دولة الإمارات العربية المتحدة المعمول بها في إمارة دبي.

  • CheckedThe Arabic and English texts say the same thing.
An illustration of the kind of review such a system produces, and not a screenshot of a finished product. It shows what was found, the rule behind the finding and the suggested change. The decision rests with the lawyer.

We build private AI systems that give your contracts, trade documents and client files a careful first reading before a lawyer opens them. Each document is checked against your own rules and templates, in Arabic and English, and every point is tied to the page it comes from. The system runs on servers that you control and follows the law that governs your contracts, whether UAE mainland law, DIFC or ADGM, and your lawyers approve its work. We are a project team: each system is built for one client, and we do not sell a ready-made product.

A first conversation carries no obligation, and we would be pleased to meet you in person in Dubai.

Prefer your own email?

You can also write to us from your own mail program, and one of us will reply personally within one working day. The button opens a new message in your mail program, addressed to us, with the subject filled in and a few headings for the things we need to know. If a mutual acquaintance suggested that we speak, please do mention them.

Send a project enquiry

projects@algomaschine.tech

The first step

A First Review of one process

We take one process on which your team spends the most reading time, such as incoming supplier contracts, letters of credit before presentation or client onboarding files. We study it from start to finish, using a sample of your past documents.

You then receive a written assessment. It sets out what is worth automating, what should stay with people, where the risks lie, and what a system would need to read and check.

Your partA first meeting, a sample of past documents, and time with the person who runs the process.

Request a First Review

The button opens a prepared email. You can also write to projects@algomaschine.tech.

What we are
An organised project team. We design and build a system for each client and do not sell packaged software.
Where it runs
On your own servers, or in a private cloud environment that you control. The data can stay in the UAE where that is required.
Who decides
Your lawyers review and approve every result before anyone relies on it.
What it follows
Your own rules and templates, and the law that governs your contracts.

What we are

We are a project team. We do not offer a ready-made product. For each client we design a system, build it, test it on that client’s own documents and hand it over.

We work as one organised team in which each of us has a defined role. We do the work ourselves and see it through to a finished, working result. You would be working with one team, and with the same people from start to finish, rather than with contractors brought together for a single job.

AlgoMaschine Intelligence is a new team, and what we bring is experience. Each of us has designed, built or run systems of this kind in earlier roles, for banks, insurers, retailers and trading firms. A system for your company would be built new, together with your lawyers and your IT team.

From our earlier work

  • Eduard designed and built a proprietary AI-driven marketplace for trade finance.
  • Igor designed a system that answers plain-language requests for reports, built on AI language models that can run on a company’s own servers.
  • Igor designed a voice assistant for a brokerage, based on a language model given further training on financial texts.
  • Igor ran a document-recognition pipeline that read 50,000 to 70,000 supplier invoice pages a day, with staff checking the results.
  • Dmitry builds and runs AI-powered export and import intelligence platforms that help manufacturers to find target markets, buyers, suppliers and trade exhibitions.
  • Eduard builds machine-learning models for financial markets and tests them until they hold up on data they have never seen.
  • Eduard has built small AI agents: one reads program code and reports on it, and another turns voice messages into mind maps.

Anna and Dmitry bring the practice that these systems must serve: legal opinions and due diligence on cross-border transactions, and the everyday documents of international logistics and trade.

What we answer for

We take responsibility for what we build. Before work begins we agree with you, in writing, what will be delivered and how it will be tested, and we stand behind the result.

Much of our earlier work was done for banks, insurers and brokerages, where systems are audited and mistakes are costly. That is the standard we are used to being held to, and the one we hold ourselves to.

Our name and our reputation matter to us. We work openly, under our own names, and we would like each client to be glad to recommend us to the next.

Where such a system helps

Legal teams spend much of their week reading the same kinds of documents. These are the kinds of work for which we can build a system. Beside each one is an illustration of the result such a system would produce.

Trading and commodity companies

Letters of credit, bills of lading, invoices and sale contracts can be checked against one another before they go to the bank, so that a discrepancy is found in your own office and corrected in good time.

What you would measureHow many presentations reach the bank clean, and how often payment waits for an amendment.

For a First ReviewA sample of your past letter of credit sets.

Check of a letter of credit set Illustration
Letter of credit, field 44E
Port of loading: Jebel Ali
Bill of lading
Port of loading: Khalifa Port
  • FoundThe bill of lading names Khalifa Port. The credit requires shipment from Jebel Ali.
  • RuleUCP 600, sub-article 14(d): the documents must not conflict with the credit.
  • SuggestsAsk the carrier for a corrected bill of lading, or ask for the credit to be amended, before presentation.
  • DecidesYour trade team. Nothing goes to the bank without approval.

Family groups and holding companies

Shareholders’ agreements, board resolutions and company records from several countries can be kept in order and searched in private. Ask who may sign for a subsidiary, and the answer comes with the resolution and the page behind it. The system remains within the family office and answers only to it.

What you would measureHow quickly a question about the group’s companies is answered, with the document behind the answer.

For a First ReviewA sample of the records of one company in the group.

Question to the company records Illustration
Asked
Who may sign a share transfer for the holding company?
  • AnswerAny two directors, acting jointly.
  • SourceBoard resolution of 14 March 2023, paragraph 4, page 2.
  • CheckedNo later resolution in the records changes this.
  • ConfirmsYour counsel, before anyone relies on the answer.

Banks, fintech and regulated firms

Client onboarding files, know-your-customer (KYC) records and policy updates can be read against the rules that apply to you, with each conclusion linked to the document on which it is based.

What you would measureThe time an analyst spends on each file, and how many files come back from second-line review.

For a First ReviewA sample of closed onboarding files.

Review of an onboarding file Illustration
Register of shareholders
Issued eleven months before the application.
  • FoundThe register in the file is eleven months old. Your policy asks for one issued within the last three months.
  • RuleYour onboarding policy, section 5.3
  • SuggestsAsk the client for a current register before the file goes to review.
  • DecidesYour compliance analyst.

The same approach can serve shipping, insurance and real estate teams, for charter parties, claim files, policies and sale agreements.

The people you will work with

One team carries each project from the first meeting to the handover. We do the work ourselves, and you deal with us directly. We meet our clients in person in Dubai.

Portrait of Eduard

Eduard

Team Lead & Head of Technology

Eduard leads the team and is responsible for the building and testing of the AI. He has worked on the trading systems of banks and brokers since 2007, including at Credit Suisse in Zurich and at Swissquote. He spent many years in technical support for electronic trading, including second-line support at Swissquote, on call and on site on the trading floor. That is where he learned to get things fixed on the spot and to prevent incidents before they happen. Since 2018 he has run AlgoMaschine, where he builds machine-learning models and tests them rigorously before anyone relies on them. He has also designed and built a proprietary AI-driven marketplace for trade finance. He holds every system to the same rule: it is tested on documents that were kept aside while it was being built, and it is accepted only if it handles those correctly.

Practice
Large language models, retrieval-augmented generation and AI agents, machine-learning pipelines, model validation
Earlier work
Application support for electronic trading at Credit Suisse and Swissquote. Head of IT at investment and brokerage firms.
Education
BSc Computer Science, University of Nicosia
Certification
ITIL v3 Foundation, Microsoft Certified Professional (MCP), Octalysis (gamification)
Languages
English, Russian, French, Greek
Portrait of Anna

Anna

Invited Legal Expert & AI Enthusiast

Anna is an accomplished legal professional with nine years of experience as both an external consultant and an in-house manager. Since beginning her career in 2017, she has built deep expertise in international tax structuring and planning, legal documentation review, and a wide range of corporate taxation issues, along with advising on certain court procedures, foreign exchange and customs matters. Having worked both inside companies and in international consulting, she understands business pain points from both sides, a rare dual perspective that allows her to combine legal rigour with commercial pragmatism. As an AI enthusiast, she examines the systems we build with the discernment of a savvy auditor and challenges them like opposing counsel in a courtroom. Her blend of legal, tax and financial insight makes her the person our team turns to for the rules each system must follow, and her gift for working with people makes her a bridge between the technical team and the real business needs of our clients.

Practice
M&A tax structuring, transfer pricing, controlled foreign companies, currency control and customs matters, tax review of legal documents
Education
Master’s degree in tax law with honours, Lomonosov Moscow State University. Bachelor’s degree in law with honours.
Speaking
International conference on the legal framework for trade within the Shanghai Cooperation Organization, Shanghai, 2025
Languages
English, Russian, French
Portrait of Igor

Igor

IT Architect

Igor designs how each system is built and kept secure. He has more than twenty years’ experience of designing large business systems for banks, insurers, retailers and financial technology companies. His recent work includes a system that lets staff ask for reports in plain language, built on AI models that can run on a client’s own servers, and a redesign of the access rules at a large retail bank, so that auditors could verify who is able to see what. He makes sure that each system is kept separate, works with the software you already use and is monitored once it is in service.

Practice
Solution architecture, enterprise integration, multi-tenant and high-load systems, LLM and RAG architecture
Earlier work
A document-recognition pipeline handling 50,000 to 70,000 invoice pages a day, with human review of the results
Education
BSc Business Administration (Management Information Systems), University of Nicosia. MSc Psychology.
Certification
Microsoft Certified Professional (MCP) · Microsoft Certified Systems Engineer (MCSE) · Microsoft Certified Database Administrator (MCDBA) · Microsoft Certified Trainer (MCT)
Languages
English, Russian, French, Greek
Portrait of Dmitry

Dmitry

International Trade and Logistics Advisor

Dmitry is our specialist in international logistics and trade, with more than seventeen years in the field. He began in ship management, where he coordinated up to twelve bulk carriers at a time and some 4.5 million tonnes of cargo through ports in the United States, Brazil, India, China and South-East Asia. He has since run import and export departments for industrial and chemical goods, covering the whole chain from supplier contracts and certification to customs and delivery. His current projects are AI-powered export and import intelligence platforms for manufacturers, which he designs, builds and runs. Each one brings market analysis, buyer and supplier research, exhibition research and a step-by-step trade strategy into a single roadmap, and is built with language models, AI agents, Python, APIs and databases. He makes sure that the systems we build for trading companies reflect how trade documents work in practice.

Practice
International supply chains, sourcing from China, customs clearance and tariff classification, Incoterms, freight and logistics, product certification
AI work
Export and import intelligence platforms built with LLMs, AI agents, Python, APIs, databases and web applications
Earlier work
Operations manager at a ship management company: bulk carrier operations, port agents, insurance and claims
Education
BSc Management Information Systems, University of Nicosia. Course in time chartering, Navistar Shipping.
Languages
English, Russian, Greek

How it differs from an ordinary AI tool

Most AI tools for lawyers work in a similar way. You type or paste your text into a website or an app, and it is sent over the internet to a large AI model operated by another company. A single model serves every customer. That arrangement suits everyday writing. A shareholders’ agreement or a financing file calls for more: certainty about where the text goes, and a clear account of how each answer was reached.

The systems we build run on your own servers and are set up for your work alone. This is how we design each one.

An ordinary AI tool

YOUR COMPANY Your documents contracts and files sent over the internet The AI tool a website or an app passed on Public AI model one model shared by every customer
Your documents leave your company and are read by an AI model that serves everyone.

A private system with a control layer

YOUR OWN SERVERS Your documents templates, rules and past contracts CONTROL LAYER (HARNESS) Finds the documents Applies your rules The AI model Checks the answer draft Your lawyer approves every step Full record Nothing leaves your servers.
Your documents, the AI and the record of its work stay on your own servers, and a lawyer approves the result.

Your documents stay with you

In a system we design, the AI, your documents and the record of what was done are all held on servers that you control, whether your own or a private cloud environment to which only you have access. If the data has to remain in the UAE, it remains in the UAE. Nothing is sent to an outside AI company unless you decide that it should be.

For your IT teamSelf-hosted language models, on your own hardware or in your own cloud tenancy. No calls to external AI services unless you permit them.

Shaped around the way you work

We begin with your own contract templates, your standard clauses and your approval rules. The system is taught which terms your company accepts, which it declines and who has to sign off. It works in the languages of your documents, including contracts that set an Arabic and an English text side by side.

For your IT teamYour templates, clause library and contract rules become the system’s reference library and rule set. Your data is not used to train a shared model.

The AI is checked at every step

An AI model on its own can sound certain and still be wrong, so we build a control layer around it. This layer finds the right documents for the AI to read, applies your rules, checks each statement against the document it came from, keeps a record of every step and pauses for a lawyer’s approval. Engineers call it a harness. Most of our effort goes into this layer, because it is where a wrong answer is caught before anyone relies on it. The result is an answer that your lawyer can check against the document itself.

For your IT teamRetrieval-augmented generation over your document store, rule checks on every output, answers tied to their source passages, a full audit log and human approval before release.

Law, systems, machine learning and trade in one team

This work calls for a lawyer, a systems architect and a machine-learning engineer working on the same problem. For trading companies it also calls for someone who knows international logistics and trade from the inside, and our team includes one. We know the paperwork of international trade from practice: supplier contracts, customs declarations, shipping documents and currency rules.

How we work with you, step by step

Every system is a project in its own right. We work in six steps, at a pace that suits your organisation, and each step ends with something tangible: a document, a design or a working system.

  1. Introduction

    We meet, in person in Dubai wherever possible. You tell us where your legal work takes the most time, and we sign a confidentiality agreement before we see any document.

    You receiveA signed confidentiality agreement

    Your partOne meeting, and your signature on the confidentiality agreement

  2. First Review

    We study one of your processes from start to finish, using a sample of your past documents. We then set out in plain terms what is worth automating, what is better left to people, and where the risks lie.

    You receiveA written assessment, which is yours to keep whether or not we continue

    Your partA sample of past documents, and time with the person who runs the process

  3. Design

    Our architect and your IT team agree where the system will run, what it may read and who approves its work. Your lawyers set the rules it must follow, and our legal advisor helps us to build them into the system.

    You receiveA design document covering where the system runs, the data it uses and the risks

    Your partYour IT team and your lawyers take part in the design

  4. Build

    We assemble the system. Your templates and past documents become its library, your contract rules become its checklist, and every action it takes is recorded.

    You receiveA working system, tried out on test documents

    Your partYour templates, your rules and answers to our questions

  5. Testing

    We test the system on your own past documents. Your lawyers mark its work and we correct it until they are satisfied. Its security is tested at the same time.

    You receiveA test report signed off by your team

    Your partYour lawyers mark the system’s work and sign off the report

  6. Handover and support

    We install the system on your servers, train the people who will use it and prepare the documents your auditors will ask for. We remain available afterwards, because laws and templates change and the system has to keep pace with them.

    You receiveA live system, its documentation and continuing support

    Your partThe people who will use the system attend the training

The security and privacy standards we build to

Auditors assess how a company protects its data by reference to published standards and laws. The UAE has set clear rules for personal data and for the use of AI, both on the mainland and in its financial centres, and we design to them from the first meeting.

Information security and privacy

  • ISO/IEC 27001:2022Keeping information secure
  • ISO/IEC 27701:2025Protecting personal data
  • SOC 2Independent audit of security controls

AI governance

  • ISO/IEC 42001:2023Managing AI responsibly
  • EU AI ActEuropean law on AI, Regulation (EU) 2024/1689
  • OWASP LLM Top 10Known security risks of AI language tools

Data protection law

  • UAE PDPLUAE personal data law, Federal Decree-Law No. 45 of 2021
  • DIFC Data Protection LawDIFC Law No. 5 of 2020, with Regulation 10 on AI
  • ADGM DPR 2021Data protection rules of Abu Dhabi Global Market

We build each system to these standards and record how every requirement is met. The certificate itself is issued to the organisation that operates the system, by an independent certification body. Our part is to give your auditor a system, and the supporting evidence, ready for that review.

Where your work extends to Europe, the same design covers the GDPR and the EU AI Act. For work connected with court proceedings we follow the DIFC Courts’ Practical Guidance Note No. 2 of 2023 on the use of generative AI.

Questions you may have

Will our documents leave our premises?
A system we build runs on your own servers, or in a private cloud environment that only you control, and in the UAE if that is required. It makes no calls to outside AI services unless you decide that it should. For a First Review you choose which past documents to share with us under the confidentiality agreement, and if you prefer we can read them at your office.
What if the AI is wrong?
It can be, and so we build a control layer around it. Every statement is checked against the passage it comes from, every step is recorded, and nothing is released until one of your lawyers approves it.
Is our data used to train anyone else’s model?
No. Your templates, rules and documents become the reference library of your own system. They are not used to train a model that anyone else uses.
How much of our team’s time does this take?
A First Review needs a first meeting, a sample of past documents and time with the person who runs the process. If we go on to build a system, your IT team takes part in the design, and your lawyers mark the system’s work during testing.
How well does it handle Arabic?
A system we build works in the languages of your documents, including contracts that set an Arabic and an English text side by side, and it checks that the two say the same thing. We test it on your own bilingual documents before you rely on it.
How is the cost set?
After the First Review we propose a scope in writing: which documents, which checks, and where the system runs. The price follows from that scope and is agreed before work begins.
What happens when the law or our templates change?
We remain available after the handover, and we update the rules and the library as your templates and the applicable law change.
You are a new team. Why should we start with you?
The first step asks little of you: a confidentiality agreement, a sample of past documents and a written assessment that you keep. We would rather show you our method on your own past documents than ask you to take our word for it, and you can judge us on that before you decide anything else.