Executive Summary

In the capital markets ecosystem, data is the engine, but punctuality is the rule. Within the Risk Department of a major Corporate and Investment Bank (CIB), producing risk metrics (VaR, Limits) is a critical mission processing 6 billion data points daily.

Faced with this explosion of complexity, MARGO integrated a pioneering AI expertise cell to transform manual supervision into a predictive and autonomous system. By deploying a cutting-edge multi-agent architecture, we secured the SLE (Service Level Expectation) while restoring value to the analysts’ work.

The Strategic Challenge: Overcoming the limits of human supervision

The infrastructure digitalization project took place in a „Scaling Law” context: the volume of data had become so large that manual tracking of calculation chains was structurally impossible to perform accurately.

• Operational Risk: Any incident on the calculation chain starting at midnight threatens the delivery of indicators to regulators and traders.
• The Bottleneck: Teams of experts were mobilized on repetitive support tasks at the expense of high-value risk analysis.
• Cognitive Debt: A massive base of business procedures, difficult to mobilize instantly during critical nighttime incidents.

The Solution: Hybrid Artificial Intelligence

MARGO designed a dual strategy, combining the statistical power of traditional Machine Learning with the reasoning capabilities of Generative AI.

Anticipation: The Predictive Monitoring Engine (PSM)

Using predictive maintenance models, this tool analyzes data flows in real-time from the very beginning of the cycle.

• Innovation: It associates a probability of success with SLE compliance.
• Impact: If a weak signal indicates a potential delay, operational teams are alerted before the anomaly becomes critical.

Autonomy: Intelligent Multi-Agent System

Based on LangChain and LangGraph, this system goes beyond a simple chatbot to become a true agentic „co-pilot.”

• RAG Pattern (Retrieval-Augmented Generation): Instant access to all business documentation for ultra-fast diagnosis.
• ReAct Pattern (Reasoning & Acting): The AI does not just respond; it reasons, identifies anomalies in logs, and suggests concrete resolution steps.

Methodology: Trustworthy AI (Human-in-the-loop)

For a highly regulated financial environment, AI cannot be a „black box.” MARGO implemented three pillars of trust:

Business Impact & ROI

The implementation of MARGO’s solutions generates tangible benefits, transforming support into a performance lever.

Outlook: Toward Self-Healing

The success of the 2024 MVPs paves the way for massive industrialization planned for 2025 and 2026. The next step is to extend autonomous automation capabilities to move toward a „Self-healing” infrastructure, capable of repairing minor incidents without human intervention.

Context: Why has anticipating demand become strategic?

Qubit Pharmaceuticals, a fast-growing innovative start-up, is at the forefront of chemistry research and numerical simulation, with an ambitious goal: to accelerate the discovery of new molecules and generate drug candidates from complex molecular data. Qubit Pharmaceuticals’ Machine Learning team, composed of specialized researchers and developers, called upon MARGO to develop and deploy Machine Learning algorithms (Xgboost, Lightgbm, Catboost, GNN, NN) to optimize and automate the molecular discovery process.

---

Challenges: Optimizing molecular research to become a leader in Medical Innovation

The main challenge for Qubit Pharmaceuticals is to improve the efficiency of the molecule generation process by making exploration faster, more exhaustive, and better guided. This involves reducing the need for costly laboratory tests and enabling broader and more precise molecular exploration.

The objectives are as follows:

Optimize the new molecule generation tool through Machine Learning models (Xgboost, Lightgbm, Catboost, GNN, NN).

Explore molecular space in a more exhaustive, rapid, and targeted manner.

Develop a plug-and-play tool accessible to all research teams for large-scale adoption.

Metrics:

1. Convergence speed: Measurement of the time required for models to reach an optimal solution. Improving this metric has increased the efficiency of the molecule generation process.
2. Exploration diversity: Key indicator for measuring the variety of generated molecules. The goal is to find a balance between exhaustive and targeted exploration of molecules.
3. Repeatability: Measurement of the ability to reproduce the results obtained, ensuring the robustness and reliability of the developed models.
4. Chemical prediction accuracy: Evaluation of the models’ ability to correctly predict the chemical properties of molecules, an essential criterion for optimizing the drug-candidate selection process.

These metrics make it possible to track the effectiveness of machine learning models and ensure that results align with the project’s objectives.

---

MARGO’s Response: Optimization of molecular exploration and development of Machine Learning algorithms

MARGO deployed a solution integrating machine learning models, such as Graph Neural Networks (GNN) and Reinforcement Learning (RL) approaches, to more effectively predict the chemical properties of molecules.

From data to discovery: tailor-made support

Working closely with Qubit Pharmaceuticals’ internal teams, MARGO has:

• Optimized molecular exploration by combining machine learning techniques and graph theory to model molecules and accelerate their generation.
• Implemented an evolving internal ML package, allowing for rapid and smooth integration of new discoveries into the research process.
• Created quality metrics to measure convergence speed, exploration diversity, and repeatability, thereby guaranteeing the efficiency and reliability of the discovery process.

Innovation at the heart of the project

This project perfectly illustrates MARGO’s ability to integrate cutting-edge technologies to meet complex challenges. The use of Graph Neural Networks for molecular exploration, a new and promising approach, makes it possible to fully exploit the potential of molecular data by refining predictions for new viable molecules.

ROI

MARGO’s intervention generated significant benefits:

Conclusion: Qubit Pharmaceuticals has reached a key milestone in the research of new molecules

MARGO’s AI expertise has allowed Qubit Pharmaceuticals to reach an important milestone in optimizing its processes for generating new molecules. This molecule generation project, now optimized through machine learning methods, has become a true strategic lever for accelerating scientific discoveries.

The molecule generation tool is now ready to be deployed on a large scale and will be accessible to all Qubit teams. This project is part of a broader vision for innovation, with potential new applications in quantum physics, chemistry, and medicine.

Context: Why is anticipating demand now strategic?

In the retail and distribution sector, logistics and supply chain management play a crucial role in the overall performance of companies.

The ability to anticipate demand has thus become a key factor for competitiveness.

For Saint-Gobain, a leader in the distribution of construction materials, operational efficiency relies on precise anticipation of future demand.

• Stockouts harm customer satisfaction.
• Costly overstocks directly impact profitability.

Faced with these challenges, Saint-Gobain is modernizing its forecasting processes to increase accuracy and efficiency.

Operational Challenges: Improving sales forecasting and product availability

The challenge is to provide demand forecasts for each item that are more reliable and accurate than those generated by existing business tools.

This objective is twofold:

Metrics used to evaluate prediction performance and improvement

1. Bias

• Measurement of the average overestimation or underestimation of forecasts compared to actual values.
• Example: A reduction from +5.8 units (existing) to +2.2 units (model co-developed with MARGO).

2. WAPE (Weighted Absolute Percentage Error)

• A weighted error measurement used to calculate forecast accuracy while accounting for the relative importance of each product or category.
• Example: An improvement in accuracy from 44% (existing) to 38% (model co-developed with MARGO).

MARGO’s Response: Forecasts optimized by machine learning

MARGO developed a custom machine learning solution, integrated directly into Saint-Gobain’s business tools.

This solution relies on advanced predictive models capable of providing more accurate forecasts than those generated by existing algorithms.

Main stages of the intervention:

1. Historical data analysis:

Evaluation of the existing system and identification of biases in current forecasts.

2. Construction of predictive models:

Implementation of machine learning algorithms capable of better interpreting historical data to produce more reliable forecasts.

3. Integration into business tools:

Predictions from the machine learning models are integrated directly into the replenishment systems, enabling flow automation and adjustable forecast maintenance by pilots.

4. Monitoring and validation:

Comparison of new model performance with existing forecasts to validate their effectiveness on key metrics such as WAPE (Weighted Absolute Percentage Error).

ROI

MARGO’s intervention generated significant benefits:

In Conclusion: Optimized logistics for sustainable performance

Thanks to MARGO’s expertise, Saint-Gobain has transformed its forecasting processes by leveraging advanced technologies.

This collaboration has optimized storage costs and strengthened product availability, a crucial element in maintaining customer satisfaction.

By relying on cutting-edge machine learning technologies, MARGO has enabled Saint-Gobain to reach a new milestone in managing its logistics flows, while offering a sustainable competitive advantage in a constantly evolving sector.

In an article published by Le Figaro, the La Défense district is described as a business hub in full renewal, now ranked fourth in the world and boasting an 85% occupancy rate. Far from premature announcements about the end of the office, Europe’s leading business district confirms its attractiveness thanks to its centrality, the quality of its offering, and the richness of its services.

MARGO’s move to La Défense is fully in line with this dynamic. It is within this strategic context that we chose to set up our new headquarters on the 27th floor of the Aurore Tower, thereby asserting a clear vision of the hybrid office as a lever for performance and influence.

The office as a strategic lever

The debate pitting remote work against office presence has long polarized positions. Our conviction is different: the office remains a strategic lever, provided it is designed as a space for experience and convergence.

The office is no longer thought of as a simple place of production, but as a vector for experience, innovation, and influence. Anne-Sophie Autuly, Head of Branding & Communication

We designed this new headquarters as an extension of our culture: performance, high standards, and team spirit. Modular collaborative spaces, an auditorium for technical conferences, an event rooftop, and premium services all contribute to creating a stimulating environment, conducive to learning, co-construction, and skill development.

Being part of a strong territorial dynamic

The renewal of La Défense is based on structural criteria—centrality, quality of offering, and services—that meet the new expectations of both companies and talents.

By establishing ourselves at the heart of this strategic district, we are choosing an ambitious hybrid model: an office designed as a point of convergence between employees, clients, and partners.

More than just a move, this installation reflects a conviction: in a demanding technological environment, the workplace remains a strategic tool at the service of collective performance and sustainable influence.

The structuring of MARGO’s European architecture in Poland marks a strategic milestone in the consolidation of its continental model. In an evolving European technological landscape, consulting firms must now deploy integrated execution capabilities on a European scale. Poland is thus establishing itself as a key pillar of this architecture, serving a more coherent, robust organization resolutely focused on sustainable performance.

With MARGO Polska, the group is strengthening the architecture of its European organization. The challenge goes beyond simple international development: it is about orchestrating an integrated model capable of aligning technical expertise, operational governance, and strategic consistency.

The structuring process underway in Poland is fully part of the ONE MARGO vision: a unified organization designed to streamline interactions between the Excellence, Business Development, and Talent hubs, and to guarantee homogeneous standards across the group.

This positioning responds to a profound market transformation. Companies are no longer just looking for distributed technical capabilities, but for robust operational architectures capable of supporting high-complexity projects within a secure and structured framework.

In an article published by Solutions Numériques, Alexandre Le Mée , Country Manager of MARGO Polska, shares his feedback on the evolution of nearshore and its growing role in structuring engineering models.

As technology projects become increasingly complex, nearshore can no longer be reduced to a simple logic of cost optimization. It is becoming a strategic lever, capable of supporting distributed architectures, reinforced regulatory requirements, and large-scale execution challenges.

From a capacity-based model to integrated engineering

For Alexandre Le Mée, the performance of a nearshore setup relies primarily on organizational consistency. The challenge is not to export skills, but to fully integrate teams into a unified governance structure.

The model developed in Poland follows this logic:

• alignment with the group’s methodological standards,
• joint project management,
• operational continuity between French and Polish teams,
• homogeneous technical requirements.

A long-term vision

This feedback highlights a strong conviction: the success of a nearshore model relies on a long-term vision, based on investment in talent, the consolidation of expertise, and integrated governance.

By strengthening its presence in Poland with this perspective, MARGO continues to develop an operational model capable of combining performance, risk management, and technological excellence.

A recognized expert in High-Complexity IT, MARGO officially joins the French-Polish Chamber of Commerce and Industry (CCIFP). This strategic move supports the ambition of our „One MARGO” model and our vision of technological excellence on a European scale.

Long-term roots within a fast-growing ecosystem

As 2026 promises to be a pivotal year for Poland, with a projected growth of +3.5% and massive technological investments, MARGO confirms its commitment to being at the forefront of this momentum.

Recent discussions at the French Embassy have confirmed a total convergence between our high-end expertise and the country’s development ambitions. By joining the CCIFP, we are strengthening our ties with local economic players to transform these perspectives into industrial realities.

Poland, a strategic hub for the „Premium Nearshore” model

For 20 years, MARGO has been supporting companies in their most critical challenges: Data & AI, Cloud Transformation, advanced software engineering, and financial markets. In a context of high talent pressure, we have distinguished ourselves through a unique model.

Far from traditional outsourcing schemes, MARGO Polska, created in 2017, has become a true strategic pillar. Our Warsaw-based teams are autonomous and capable of designing, managing, and industrializing large-scale projects from end to end.

The „ONE MARGO” approach: the best of both worlds

This strategy is driven by our unified organization, which guarantees our clients an international impact without compromise: