Andre Borie

Case studies

A selection of my recent, high-impact projects, in a nice business-friendly format.

Verisk Maplecroft

Verisk Maplecroft, a leading provider of geographic & industry risk data, approached me for assistance with their data platform that allows clients to analyze their assets' risk exposure across multiple indices. Their platform needed to evolve to meet growing demand as their client base and data offerings expanded.

AREA - Asset Risk Exposure Analytics

One of my major contributions was the complete specification and delivery of AREA (Asset Risk Exposure Analytics), a transformative platform that gives clients full visibility into their portfolio's risk and sustainability profile. This sophisticated solution combines the locations of millions of corporate assets with industry-leading geospatial risk data across many individual risk indices mapped to subnational levels.

AREA was specifically developed to reveal hidden vulnerabilities and strengths in the global operations of publicly listed companies worldwide, providing an entirely new way of assessing organizational sustainability and resilience. The key advantage is that clients can assess companies against all these risks in one place using a consistent scoring framework. For added granularity, 50 of the indices also incorporate industry-specific risk data to provide a weighted, sector-by-sector view of risk.

The technical challenge was designing a system capable of making billions of pre-calculated risk scores available for instant querying by customers worldwide. Working closely with business stakeholders, I gathered requirements while carefully considering infrastructure capabilities and data source constraints.

The final deliverable was a Django-based application deployed on AWS, with PostgreSQL for operational data and Redshift for score data, shared authentication with existing applications, and Okta integration. Initially, an entirely self-hostable prototype was developed using SQLite as the underlying data store which offered adequate performance and infinite horizontal scalability, but Redshift was ultimately chosen by the client due to its managed nature.

I also designed and implemented the ETL pipeline necessary to prepare and import the scores data, creating implementations that worked with their chosen data storage solution while providing flexibility for future growth.

GRiD Performance Optimization

When faced with a situation where the scoring system for GRiD (Global Risk Dashboard) struggled to process the growing volume of location data against new risk indices, I developed and deployed a replacement system in a single day to meet their deadline and complete the release of the new indices without disruption to clients.

GRiD, Maplecroft's flagship global risk intelligence platform, combines over 190 risk issues with expert analysis to provide a comprehensive web-based source of global risk intelligence. The platform helps clients quickly identify and understand risks worldwide to make better business decisions.

My optimization solution featured an adaptive approach that maximized scoring service utilization without overwhelming it and dynamically reacting to scaling events. This solution proved highly effective, resolving their immediate challenge and providing a sustainable foundation for future growth as GRiD continued to expand its comprehensive risk index portfolio.

I also provided architectural guidance on potential infrastructure improvements to eliminate bottlenecks and scale more effectively, considering both performance and cost implications for this business-critical platform.

GRiD Platform Enhancement

Throughout the engagement, I maintained and extended the REST API of the GRiD platform, implementing new endpoints using Django REST Framework while modernizing legacy code. This ongoing work significantly improved code quality, performance, and maintainability of this flagship product.

Key contributions to the GRiD platform included:

Throughout this 2-year engagement, I worked across multiple aspects of Verisk's technology stack while serving as a technical consultant to their team. This included providing guidance to developers across multiple projects, consulting with infrastructure engineers on performance concerns, and helping the team navigate cloud infrastructure challenges.

Coconut

Coconut reached out for urgent assistance with their cloud-hosted Django application powering an electronic money service including debit card issuance and innovative, real-time bookkeeping features aimed at small business customers.

The primary objective was to support them during a pivot to an Open Banking model where the product's bookkeeping features could be applied to any third-party bank account, with the aim of eventually phasing out the e-money service. Successful and timely delivery was paramount as this was the main selling point behind their latest funding round.

As the technical consultant working directly with Coconut's leadership and development team, I functioned as an interim technical advisor at the CTO level during this critical transition phase. I provided strategic direction, made key architectural decisions, and guided implementation while they searched for a permanent CTO, all while developing a new & robust Open Banking integration.

Open Banking integration rebuild

Their existing Open Banking integration, while a valiant effort, had major connection state management shortcomings where bank account connections would get stuck in a broken state, prompting customer complaints, significant operational overhead (the connections would need to be manually reset by staff) and decreased customer trust.

The solution was a brand new Open Banking integration with their partner TrueLayer, featuring robust state management and tolerance to bank/partner outages with automatic recovery. This significantly decreased Open Banking-related customer queries and eliminated manual cleanup work as the system was able to self-heal after transient issues. The new solution also gracefully handles the regular reauthentication & consent requirements imposed by PSD2.

Open Banking data cleanup

The next challenge was to make the system tolerant of the varying quality of the transaction data and provide a great user experience regardless of whom they are banking with. A transaction matching algorithm was developed using various heuristics, which was then back-tested on historical data to confirm accuracy. The end result is a system that is able to tolerate changes to historical transactions' descriptors and/or IDs - a common challenge with established banks' data and something competing products often struggle with and may import duplicate transactions instead. This effort also involved specifying and coordinating changes to the mobile clients' internal data model.

Beyond technical implementation, I also contributed significantly to product decisions around the Open Banking connection management & onboarding flows, creating UI mockups and collaborating with the product & design teams to ensure a seamless user experience that balanced technical constraints with usability.

Throughout the engagement, Coconut - like all other clients - also benefited from day-to-day technical assistance, release engineering, infrastructure maintenance, monitoring and disaster recovery, including out-of-hours assistance when necessary.

Babylon Health

Babylon Health (since acquired by eMed) approached me with a critical challenge related to their health risk assessment and advisory application. This app analyzes user-provided health information and risk factors to predict potential health risks and provide personalized preventive recommendations – for example, identifying increased lung cancer risk for smokers and recommending smoking cessation. As a regulated medical device in several jurisdictions, this application required rigorous clinical safety testing to meet regulatory requirements.

The primary issue was that their testing process relied on clinicians manually executing "case cards" – scenarios involving model patients with specific risk factors and expected health recommendations. For instance, ensuring that a 45-year-old male smoker with a family history of heart disease would receive appropriate cardiovascular risk warnings and lifestyle modification advice. This manual process was not only time-consuming but also prone to human error, creating both operational inefficiency and potential compliance risks.

As the technical consultant on this 4-month engagement, I worked closely with Babylon's clinical and technical teams to design and develop a solution that would automate this critical testing process while maintaining the clinical rigor required for a regulated medical device.

The solution was a sophisticated web application built within their existing Django framework that served as a repository for these model patient scenarios. The application was designed to simulate a mobile client by making the same API requests that the actual mobile app would make, effectively "playing back" the patient risk profiles against the live recommendation engine.

A key feature of the system was its ability to automatically compare test results against expected outcomes, clearly flagging any deviations that might indicate clinical safety issues. This allowed clinicians to focus their expertise on writing new case cards and analyzing discrepancies rather than on the mechanical aspects of test execution.

The application also enabled scheduled automated testing, ensuring that any changes to the recommendation engine could be immediately evaluated against the entire library of clinical test cases. This significantly strengthened their quality assurance process and provided better documentation for regulatory compliance.

Beyond building the new functionality, I refactored portions of their existing codebase to align with Django best practices, improving performance, security, and data integrity. These improvements not only supported the new testing functionality but also enhanced the overall application architecture.

Throughout the engagement, I provided guidance and support to Babylon's development team regarding Django best practices, helping to build their internal capabilities. This knowledge transfer was particularly valuable as it enabled their team to maintain and extend the testing platform after my departure.

The resulting system transformed the product's clinical safety testing process, reducing test execution time from hours to seconds while simultaneously improving accuracy and documentation. Tests that previously required manual execution by highly skilled clinicians could now be run automatically, freeing these valuable team members to focus on analyzing results and improving the product's clinical safety.

This solution not only improved operational efficiency but also strengthened Babylon's regulatory compliance position by providing more comprehensive, consistent, and well-documented testing of their medical device software.