ALM core
Barycentre
Repricing gaps, NII and EVE sensitivity, and IRRBB measurement under Basel standards.
The centre of mass around which a system balances.
Bring us a regulation, a model, or a question no vendor has productised. Our lab turns it into working software — researched, designed with AI, engineered to bank-grade standards.
R&D sits at the heart of the company. Every product starts as a question our lab sets out to answer.
Agents with defined roles, permissions and limits — and a human expert approving every material decision.
Audit trails, model governance and regulator-ready outputs from day one. Built to pass scrutiny, not just demos.
Like the AI assistants your teams already use — but built for a regulated bank. Every agent has a role, every action is permissioned, and every answer passes through four steps you can see.
Each agent has a defined role, data permissions and limits — like a well-run team.
Material decisions stop for sign-off by the right person, every time.
Every prompt, data access, calculation and output is recorded.
Runs on private models, on-premises or in-country. Your data never trains shared models.
Your policies and regulatory rules are encoded as constraints agents cannot cross.
Every number traces back to source data, model version and assumptions.
Where we study how AI can be trusted inside regulated institutions.
The workspace and agent platform every product runs on.
Suites of specialist agents — starting with the balance sheet.
Agentic intelligence for the balance sheet. ALM, IRRBB, FTP, IFRS 9 and behavioural modelling — each module a team of governed agents, named after the physics of balance.
Repricing gaps, NII and EVE sensitivity, and IRRBB measurement under Basel standards.
Non-maturity deposits, prepayments and early withdrawals, modelled from your own customer history.
Staging, PD, LGD and EAD, and expected credit loss with full traceability from data to disclosure.
Forward-looking balance-sheet projections and scenario stress testing, from base case to severe.
Optimise the balance sheet against capital, liquidity and funds-transfer-pricing constraints.
Live dashboards, board packs and regulatory returns that explain themselves.
We are a research lab first. We study how AI agents can be trusted inside regulated institutions — and turn that research into products banks can run on.
Roles, permissions, escalation and approval for teams of AI agents working inside a bank.
Smaller, specialised models that run on-premises and understand banking, risk and regulation.
Evaluation, validation and ongoing monitoring of AI — held to the same standard as any bank model.
Answers that show their working: sources, assumptions and model versions, in plain language.
Behavioural, credit and balance-sheet models designed and calibrated with agents, reviewed by experts.
Benchmarks and test harnesses that prove an agent does the job correctly — before it goes live.
Designed for the data-residency expectations of Gulf regulators. Lmar and AIFA run where your data already lives — and your data is never used to train shared models.
Installed inside your own data centre, under your security controls.
Hosted on approved local cloud regions in the UAE or the Kingdom of Saudi Arabia.
Fully isolated deployments for the most sensitive environments.
Our team has implemented balance-sheet, capital and risk systems at banks across the Middle East. We bring that experience to every engagement — with or without our software.
ALM, IRRBB, FTP, liquidity and IFRS 9 methodology, design and review.
Management and board dashboards that turn risk data into decisions.
Data pipelines from core banking and treasury into risk and reporting systems.
Automated data-quality checks and reconciliations you can evidence to audit.
Test strategy, scripts and execution for system implementations and upgrades.
Ongoing run support, model monitoring and regulatory change updates.
We start with your problem, your data and your regulator — not a template.
Our lab prototypes on your own data, quickly, with governed agents.
Validated with your risk, finance and audit teams until every number stands up.
Into production on Lmar, with documentation, handover and support.
Senior practitioners on your project, when you need them.
A defined question, a defined outcome — from blank page to working prototype on your data.
The platform and product suites with support, updates and regulatory change included.
Every sun, ours included, begins as a dark cloud: gathered, shaped, ignited. And once it shines, it hides nothing — whatever it reaches can be seen. That is how we believe AI should work inside a bank: in the open, every step visible, every answer traceable.
And the name carries our method. Every answer Lmar gives passes through four steps.
Every number starts from your own data, reconciled to source.
Specialist agents apply your approved models and assumptions.
Nothing material moves without a person signing it off.
Every answer arrives with its evidence: data, model version, assumptions, approver.
Four letters, four steps, every time. Every answer, in the light.
Tell us about the problem no vendor has productised. We’ll come back within two working days.