mmtb

MMTB  /  Applied artificial intelligence

Artificial intelligence, engineered to stay in production.

We are a boutique AI firm. We design the strategy, build the system, and stay with it until the value shows up in your numbers.

Headquarters  ·  Johannesburg
mmtb.co

The work is only finished when it runs in production, in the hands of the people who need it.

01The practice01

A boutique firm that finishes what it starts

MMTB exists to help our clients and our own people make a measurable difference through applied artificial intelligence. The discipline is simple to state and difficult to hold.

We are deliberately small. We apply modern architecture and advanced AI methods to improve how a business already operates, or to help you build an AI business of your own.

We work alongside your teams rather than around them, because the pragmatic solution is usually the one your people can carry after we leave. We do not stop at proof of concept. We stay until the system is stable, owned internally, and adding value you can audit.

Direction 01

Build AI strategies

  • Set the AI vision and roadmap with senior leadership
  • Name the real gaps in current capability
  • Identify business models that AI makes possible
  • The roadmap is a document and a build plan at once, sequenced against work we could deliver ourselves
Direction 02

Build AI capabilities

  • Build AI teams and grow the people inside them
  • Artificial intelligence, infrastructure and cloud, agile ways of working, and product deployment
  • Proprietary instruments to shorten the journey, including our AI and product maturity index
Direction 03

Build AI products

  • Full product lifecycle, from first idea to live system
  • Use case selection and a problem statement worth solving
  • Machine learning and large language models: predictive analytics, optimisation, simulation and digital twins
02Why partner with MMTB02

People who have done the work, technology that holds, delivery that lands

People

Sector and functional depth

More than twenty years of collective experience across digital, advanced analytics, and AI, matched with operating experience inside heavy industry rather than only advisory exposure to it.

Consultants who have also delivered

The team comes out of top tier management consulting and has worked across countries, functions, and industries. We do strategy and we do execution, and we would rather be judged on the second.

Resourcing that flexes

Teams are shaped around the scope, the timeline, and the problem in front of us, not around a standard staffing pyramid that has to be filled.

Technology

End to end product development

Data engineers, data scientists, visualisation specialists, and full stack developers, supported by translators who understand your process and what implementation actually costs.

Cloud and infrastructure

MLOps, DevOps, system architecture, and cloud engineering, used to deploy AI and analytics that stay reliable in mining, construction, manufacturing, and services environments.

Solutions specific to you

Real time optimisation models and condition monitoring built for your operation and integrated with the systems your people already use every day.

Delivery

Embedded collaboration

Technical leads, analytics translators, and digital consultants sit inside client teams. Alignment starts on day one rather than at the first steering committee.

No gap between strategy and execution

Strategic consultants are paired with programme managers and agile coaches, so high level goals become clear delivery roadmaps instead of rework.

Execution at scale

Agile coaches, product managers, and system architects working to structured governance, which is what keeps outcomes consistent across large multi-stakeholder programmes.

03The route03

One route, from the first question to a system in production

Each stage has its own accelerators, so that time is spent on the parts of the problem that are actually yours. Capability building and user adoption run underneath all three.

Stage 01

Digital strategy and use case ideation

  • Maturity assessment, short and long term planning
  • Use case ideation and problem definition workshops
  • Value assessment and use case prioritisation
Stage 02

Build MVPs and full production systems

  • Initial model, validation of value, agile setup, rapid experimentation
  • Production builds with a human centred approach throughout
  • Integration with the processes and systems you already run
Stage 03

Post-production support

  • Model monitoring and A/B testing framework
  • Business value confirmed after go-live, not before
  • Continuous model improvement

Impact and value

  • Start small, scale deliberately. Focused, high value use cases first, so budget exposure stays low until the proof holds.
  • Open source first. You do not carry licence fees you did not need to take on.
  • Your intellectual property. Clients keep full control to modify and build on everything we create.

People centred adoption

  • Change management embedded. Communication, stakeholder alignment, and tailored training, so the system is used and not merely delivered.
  • Knowledge transfer. Documentation, training, and handover planned from day one.
  • A business literate team. Technical depth paired with commercial judgement.

Risk managed delivery

  • Business problem first, not model first. Use cases chosen for measurable success criteria, then tested for feasibility.
  • Transparent delivery. Short sprints, frequent milestones, stakeholder demonstrations.
  • Governance and ethics. Interpretable models wherever viable, responsible AI principles applied to mitigate bias.
04Evidence04

Products built, deployed, and sustained on the plant

Mining programmes where the system stayed in service after handover, and the return the client measured against it.

0×
Return · concentrator

Increase recovery and yield, reduce waste

MMTB managed the programme and supported the client to sustain the product after deployment. Yield improved by 1 per cent.

Yield
+1%
0×
Return · smelter

Reduce energy consumption at the furnaces of a nickel smelter

Implemented at the plant and sustained after deployment. Energy consumption reduced by 3 per cent.

Energy
−3%
0×
Return · concentrator

Reduce variability by optimising weekly blends

Blending AI advisory implemented to minimise the variability of daily blends and reduce blend changes against product quality.

Variability
Reduced

Gold · Southern Africa

CIP recovery improved by 5 percentage points

  • Homogenising feed ore quality. Several mines of varying quality fed one processing site. We built models determining how the silos should be filled by train so that the ore arriving at the plant was even.
  • Improving the milling process. Models to improve mill output particle size distributions by optimising feed rates, densities, and other operating parameters.
  • CIP optimisation. Carbon usage, activation, fines management, and residence time optimised against feed quality.

Diamonds · Southern Africa

Maintenance AI, designed to be maintained

  • Predictive maintenance. We designed the strategy with the client, including sensor requirements by equipment type and the data architecture to support them, then built minimum viable products where the data was already good enough to model.
  • Work management. A framework that carries the maintenance AI requirements, so operational workflows and predictive capability stay aligned rather than drifting apart.
05Banking05

Customer 360 is the foundation of AI in banking

A unified, accurate, real time view of each customer across every product, channel, interaction, behavioural signal, and risk indicator. It is not a data lake, not a CRM, not one very large table, and not an IT project, which is the assumption that quietly kills most attempts at it.

Wave 01

Foundation

  • Clear visibility of customer data, and insight that can be acted on
  • The foundation on which the use cases sit
Wave 02

Data management

  • Churn improved by 3 to 5 per cent
  • Lifetime value up by 3 to 10 per cent
  • Satisfaction up by 1 to 10 per cent
Wave 03

Intelligence

  • Churn improved by 3 to 10 per cent
  • Lifetime value up by 3 to 50 per cent
  • Satisfaction up by 1 to 15 per cent

Ranges are indicative and drawn from comparable programmes. Actual outcomes depend on data maturity, product mix, and the adoption work that runs alongside the build.

Case · sales and marketing AI

Two per cent more sales, from a recommender built on Customer 360

Situation
A major Australian bank wanted better retention and cross-sell in a hardening market. Despite heavy digital marketing spend, conversion stayed low. Campaigns leaned on age, income, and postcode, and sales teams could not see who would respond to what.
Solution
A Customer 360 database drew each record together from sales, product, and service datasets. On top of it we built a campaign recommender that predicted churn, produced market microsegmentation, inferred unknown characteristics such as family status, and shaped campaigns for retention and cross-sell.
Impact
Sales improved by almost 2 per cent, with churn down and satisfaction up alongside it. The recommender stayed in production and continued to be retrained against live outcomes.

Case · identity fraud

Fraud caught by combining two datasets nobody had joined before

Situation
The bank's online platform saw a surge in fraudulent account openings and unauthorised transactions. Attackers used stolen identities to pass standard know your customer checks, so the controls were being satisfied by documents that were technically valid and factually false.
Solution
A machine learning detection system combining the bank's data with data held by the postal service, built inside a tightly controlled environment to minimise any leakage of personal identity information. It integrated with major shopping frameworks and returned a seamless check at the point of transaction.
Impact
Identity fraud fell significantly for both the bank and the postal service, and opened a commercial path to offer the same capability to smaller businesses.
06Leadership06
Dr Maksud Ibrahimov

Dr Maksud Ibrahimov

AI Lead Engineer and Director
PhD, Computer and Mathematical Sciences,
University of Adelaide

The technical weight of the firm sits with one person, on purpose

Maksud designs the architecture, leads the builds, and stays accountable for what happens after go-live. Fifteen years of hands-on delivery sit behind that. As Principal Data Scientist at McKinsey and Company's QuantumBlack, he ran international advanced analytics engagements from first assessment through to production deployment. Before that he founded Adaptic Solutions, one of Australia's earliest data science consultancies, and grew it from nothing to more than A$3m in revenue before it was acquired. At MMTB he sits across the advisory work and the build, which is what allows a strategy to be written as something we can be held to.

Career

2018–2023
McKinsey and Company, QuantumBlackPrincipal Data Scientist. Led international advanced analytics projects end to end.
2014–2018
Arq Group, InfoReady, Adaptic SolutionsChief Data Scientist. Built the practice to 25 professionals and A$3m+ in revenue. Telstra, NAB, AusPost, AusNet, Newcrest.
2013–2014
ABB Enterprise SoftwareProduct Owner, Scheduling and Logistics. Led a global team across ABB's mining software suite.
2008–2013
Schneider ElectricSenior Scientist. Planning, scheduling, and logistics engines for Rio Tinto, BHP, South32, Viterra, Xstrata, Incitec Pivot.

Competence

  • Machine learning and deep learning
  • Supply chain optimisation
  • Digital twin simulation
  • LLM fine-tuning and deployment
  • MLOps and production systems
  • Evolutionary and Bayesian computation
  • Enterprise AI strategy and P&L ownership

Research

  • Thesis on evolutionary algorithms for supply chain optimisation
  • 13 peer-reviewed publications in Springer, IEEE, and GECCO proceedings, more than 80 citations
  • Bronze medallist, International Olympiad in Informatics, Athens, 2004

We recommend only what we can build

Some of this work arrives as a strategy and a roadmap. Some of it arrives as a system running in your estate. The discipline is the same either way: we do not put a recommendation in front of you that our own engineers could not stand behind and deliver.

If there is a problem in your business worth putting intelligence against, send us the shape of it. We will tell you honestly what we think is possible.

Write to us

hello@mmtb.co

Headquarters

Johannesburg, South Africa

Office

Melbourne, Australia

Development hub

Philippines