Loading
Description
Squadra led the development of an Azure AI-based solution to analyze large data volumes and improve clinical outcomes.
Transcript
The Elizabeth Glaser Pediatric AIDS Foundation (EGPAF) prevents HIV transmission and treats HIV-positive patients in 12 countries heavily impacted by the virus. During the Microsoft Hack4Good hackathon held in 2025, Squadra led the development of an Azure AI-based solution to analyze large volumes of data and improve clinical outcomes, providing evidence-based insights and fostering a data-driven culture within the institution.
The solution, glAIser, uses supervised RAGs (Retrieval-Augmented Generation) and understands natural language commands, transforming them into structured queries that identify related datasets and indicators among billions of data points from over 18 million patients served. After querying, glAIser generates interactive charts, textual summaries, and practical insights.
Through rapid and facilitated reporting, the solution enables global, national, and clinic-level analyses, identifying trends, concerning patterns, and critical regions. These analyses guide resource allocation and prevention efforts, enhancing EGPAF's patient care and saving lives.
The impact perceived throughout the hackathon was so significant that, in addition to earning us recognition as global runners-up, the project didn't stop there. The partnership between EGPAF and Squadra, with full sponsorship from Microsoft Elevate, went further:
Phase 1 - Hackathon: it began at Hack4Good in Seattle (USA), where glAIser proved the potential of AI to globally impact people's lives.
Phase 2 - MVP: we developed the MVP (Minimum Viable Product), focused on analyzing only 38 local indicators through multi-agents.
Phase 3 - Scale: we moved away from the RAG-based model entirely to implement an advanced multi-agent architecture, fully integrated into the foundation's unified ecosystem, Glaser360, covering 100% of the countries where EGPAF operates.
Interpreting clinical data requires absolute precision. After all, misinterpretations can result, in EGPAF's case, in more children falling victim to vertical HIV transmission. To ensure maximum security and no hallucinations, Phase 3 had important technical milestones as pillars:
In Phase 2, the MVP, the solution was limited to reading only 38 simple, local indicators. In Phase 3, scaling, the solution began to support over 142 complex indicator combinations that are accessed directly within EGPAF's data ecosystem, Glaser360.
Furthermore, the system moved from running in a restricted test capacity to covering 100% of the countries served by the foundation's operations, all in a standardized manner within the EGPAF ecosystem.
The major technical breakthrough was the use of the LangGraph orchestration framework, which uses cyclic graph structures to create complex workflows for AI agents. This open-source framework allows for state recovery when there is an interruption in the process, offering persistence and aiding in production scaling.
Instead of a generic AI model, we created a state machine orchestrated by multiple specialized intelligent agents, where each agent fulfills a specific function in the process of treating and querying information.
To shield the operation against failures and hallucinations, we implemented a unique real-time evaluation node.
If the natural language query presents any minor inconsistency or syntax error when accessing the database in Databricks, the AI itself identifies the error, corrects it, and recalculates the execution route before displaying the result on the screen.
The new multi-agent-based solution achieved 100% accuracy during automated tests against the Golden Dataset, a massive set of real-world, highly complex data provided by EGPAF's own specialists.
The system analyzes highly complex questions, translates them into commands, and executes them in the database with an average time of 15 to 20 seconds.
The conclusion of Phase 3 generated significant repercussions among our partners, demonstrating Squadra's commitment to developing critical solutions that are innovative, secure, and truly reliable. See testimonials from those who experienced the project firsthand:
“Amazing teamwork! It’s truly inspiring to see! A heartfelt thank you to the Squadra team for your outstanding delivery, unwavering commitment, and consistently high-quality work. This has been a remarkable collaboration from start to finish.”
“Incredible teamwork! It is truly inspiring to see! A heartfelt thank you to the Squadra team for your exceptional delivery, unwavering commitment, and high-quality work. This has been an extraordinary collaboration from start to finish.”
Hugo Marques, SMB & Partner Lead at Microsoft Elevate
“This is truly impressive work, and we couldn't be more impressed with Squadra's performance and professionalism throughout this phase of the project. Seeing the solution in its current state had a pretty profound impact on me.”
“Truly impressive work! We could not be more impressed with Squadra's performance and professionalism throughout this entire phase of the project. Seeing the solution in its current state had a profound impact on me.”
Jon Drew, Account Team at Microsoft
“Thanks for the dedication and professionalism through this process.”
“Thank you for the dedication and professionalism shown throughout this process.”
Charlie Maere, Global Digital Health Director at EGPAF
The glAIser journey proves that, with the right partnership, artificial intelligence ceases to be a technical promise and becomes a robust, secure, and world-changing corporate reality.
Want to know how to apply advanced AI architectures and ensure the reliability of critical, data-driven decisions in your company? Talk to the Squadra team of specialists.
How to extract real value from generative AI and govern innovation costs to go beyond the hype?