Our Blog

How We Classified 38.6M PubMed Records: Hybrid NLP Pipelines vs LLMs

A comparison of two approaches to large scale biomedical literature classification: LLM based classification and a hybrid architecture combining named entity recognition, ontology embeddings, and disease normalization. Using 38.6 million PubMed records, we evaluated throughput, cost, ontology alignment, and traceability, finding that hybrid pipelines remain the more practical choice for production scale, ontology driven systems.

Energy Software Decision Guide: The hidden cost of bad architecture in energy systems

WEBINAR: Building end-to-end AI systems for medical imaging

This webinar walks through the full lifecycle of a medical AI system, starting from dataset creation and model training to deployment. It includes lessons learned from production systems, including performance optimization, error handling, and integration into existing workflows.

Understanding ODM in Clinical Trials, and Why It Still Matters

Clinical trials today run across a complex mix of systems. Data moves between Electronic Data Capture systems, Clinical Trial Management Systems, ePRO platforms, safety databases, and analytics environments. Without a shared structure, this exchange becomes slow, expensive, and error-prone. This is where ODM comes in.

Private AI for Regulated Industries - How to Deploy Secure LLMs Inside Your Infrastructure

Why is AI still off limits in so many regulated organizations? For healthcare, pharma, and energy teams, cloud-based LLMs raise serious concerns around data control, auditability, and compliance. This guide explains how private AI systems can run entirely inside your infrastructure, using your own documents to deliver secure, compliant insights. See how regulated teams are unlocking AI without sending data outside the firewall.

When software grows faster than its architecture - notes from our architecture session with IcePanel

As software systems grow, architecture often struggles to keep up. This article reflects on our first architecture session with IcePanel, exploring why teams lose clarity over time and how shared models, practical tooling, and real project experience can help keep systems understandable as they evolve.

Road to Microservices: Modeling Modern Systems with C4 & IcePanel

Road to Microservices is a hands-on evening event hosted by Sotex Solutions in collaboration with IcePanel, bringing together architects, developers, and tech leads to explore how the C4 Model and model-based architecture can simplify the transition to microservices. Through practical examples, live demonstrations, and open discussion, participants will learn how to maintain clarity, consistency, and scalability in complex software systems.

AI-Powered Drone Mapping – From Images to 3D Assets with up to 1m Accuracy

A single drone flight can now produce accurate, georeferenced 3D models of every asset — no manual work required. This article explores how Sotex Solutions and Beacon Power Services built an AI-powered module for CAIMS that automates object detection, coordinate computation, and 3D reconstruction with precision up to one meter, cutting survey costs by 70% and transforming asset management across industries

Why More Developers ≠ Faster Delivery

Here’s the truth: more developers don’t mean more speed. They often mean more chaos. When projects fall behind schedule, many teams turn to the most intuitive fix: add more people. On paper, it makes sense — more hands on deck should mean faster output. But in software development, the opposite is often true. Instead of accelerating delivery, scaling headcount usually slows things down and introduces risks that are far harder to recover from later.

Beyond Delivery: A True Tech Partner | Sotex Solutions

Many software projects fail not from bad code but from lack of responsibility. Sotex Solutions goes beyond delivery, owning the entire solution—from compliance and architecture to long-term scalability. With proven results in healthcare and energy, senior architects, IcePanel C4 clarity, and AI used responsibly, we build systems designed to endure, comply, and scale.