Our insights are yours

Want to discover what's been happening at Panenco this May? Our latest newsletter dives into the client work we've delivered, new technologies we're exploring, and more company updates!

In a recent SaaS project, migrating from Amazon SES to Postmark significantly improved email communication by unlocking detailed analytics on engagement, language performance, and user behavior. This transformation turned email from a basic notification tool into a strategic asset for user engagement, enabling better A/B testing and optimization across different segments.

Last month, our teammates Emma Lavens and Kristof De Troch gave a guest lecture at the computer science engineering faculty of KU Leuven on Infrastructure as Code (IaC)!

For one of our clients, a SaaS platform in the education space, we radically transformed a tedious back-office process related to image selection. We integrated Google Vision AI directly into their platform to automatically label thousands of images and enable smart search, saving their employees hours of manual work navigating complex folder structures in search for the right images.

We're growing our presence in Antwerp, being closer to some of our clients and attracting more talent! Feel free to pay us a visit at any time — at our Leuven HQ or at our hub in Antwerp. Both locations offer plenty of creative space for workshops and brainstorm sessions.

By mocking time across both the API and application levels, and simulating the execution of scheduled tasks (via GCP Scheduler, PubSub, Event Arc), we enabled testing that no longer depends on real-world time passing.

Because reading text-heavy exercises can be a real challenge for children with dyslexia, we integrated Google's Text-to-Speech AI into the entire platform, introducing a "Speech Assistant" button next to every piece of text: instructions, questions, answer options.

We recently had the pleasure of supporting ReqCon in enhancing their cloud infrastructure with a tailored Infrastructure-as-Code approach. Our short but impactful engagement focused on improving scalability, maintainability, and deployment efficiency—laying solid foundations for their next stage of growth. Always proud to empower ambitious teams with the right tech foundations

For one of our clients we recently built an agentic workflow with 10+ distinct chained LLM calls. After the initial build, the latency of the agent was large and the output accuracy insufficient. We solved this problem by implementing diligent monitoring and evaluation metrics for each step in the workflow, using Helicone (YC W23) as a useful tool in this process.

For one of our clients, we recently built an agentic pipeline that automates data ingestion from unstructured proprietary documents into structured schemas, with automated value alignment and continuous human-in-the-loop validation. As part of our tech toolkit, we've found Mastra to be a useful framework for agent development in Typescript. We recommend checking it out!

The testing phase in the Secure Software Development Life Cycle (SSDLC) is essential for addressing all identified security vulnerabilities before an application goes live. Secure coding practices and continuous testing are crucial for resilience against evolving threats, particularly when Static Application Security Testing (SAST) is part of a comprehensive SSDLC.

Working with medical text is challenging due to its complex terminology and often unstructured nature. To address these challenges, we stay continuously updated with the latest AI developments and ensure we use the best models for the task.

The growing demand for AI-powered applications has led to the development of low/no-code tools like LangFlow, which allows users to rapidly build AI applications using a drag-and-drop interface, turning ideas into quick proof of concepts. While full-scale AI applications still require traditional coding, no-code frameworks empower users to create smaller-scale prototypes with reusable components and integrated models, accelerating early-stage development.

Since incorporating Large Language Models (LLMs) into our products, we've learned that effective prompt engineering is crucial for accurate AI responses, just as well-written code ensures reliable software. Key strategies include standardizing input and output, keeping prompts concise and specific, and providing examples to enhance accuracy, all of which help optimize LLM performance and improve the user experience.

For one of our clients we’re building an enterprise SaaS product with complex procedural workflows in a highly regulated domain. The business logic contains a ton of approvals, revisions, sign-offs and other role-based features that all impact each other. In order to keep this manageable, we opted for XState as a state manager.

In order to maintain security across the REST APIs we build, we use a “divide and conquer” approach. We break down request handling into five sequential steps for nearly all endpoints:

When building user interfaces for B2B SaaS products, we typically start with creating responsive web applications, often times followed with a need to build a mobile app experience as well.

Powering up digital products with genAI services can create tremendous user value. However, three key concepts need to be taken into account when integrating AI -and in particular Large Language Models (LLMs): compliance, security and model selection.

Ensuring the security of our code repositories and cloud infrastructure is more crucial than ever. Integrating security practices into the DevOps pipeline - known as DevSecOps - is essential to maintain robust and resilient systems.

At Panenco, we've successfully implemented Infrastructure as Code (IaC) workflows, transitioning from simple YAML configurations to using Terraform and Pulumi for increased deployment speed and reliability. Pulumi, in particular, has been advantageous due to its support for familiar programming languages, comprehensive cloud resource compatibility, excellent documentation, and flexibility, enabling us to maintain structured and efficient deployment pipelines.

The widespread adoption of generative AI has led to an increase in generated text, posing challenges for database performance due to the accumulation of similar or near-duplicate strings. By using large language models to detect and merge these near-duplicates through embeddings, databases can be streamlined, reducing redundant data and enhancing overall application performance.

Using OpenAPI auto-generation to connect APIs and front-end applications addresses communication challenges and inconsistencies by generating client-side SDKs from API specifications, enhancing type safety, speeding up development, and reducing errors. This method not only provides thorough documentation but also simplifies the integration process and improves the development experience for both front-end and back-end teams.

Integrating AI services like computer vision and speech into robotics is transforming how robots perceive, interact with the world, and communicate with humans, as demonstrated by Budd-E, a remotely controlled AI-powered robot. Budd-E uses AI to visually describe its surroundings and convert text into spoken language, showcasing the potential for enhanced human-robot interaction through advanced AI and cloud technologies.

Our teammate Thomas Callens takes a trip down memory lane, reflecting on his PhD research on exoskeletons to prevent workplace injuries. He developed motion recognition algorithms for exoskeletons, achieving high accuracy and timely responses, and discovered the importance of writing robust code and systematic testing, which he now aims to deepens alongside exploring AI integration in robotics at Panenco.

We used Amazon Comprehend Medical to efficiently match patients with clinical trials by converting unstructured medical text into structured data, specifically by identifying medical entities and mapping them to ICD codes. This approach enabled us to streamline the selection process, saving costs and improving performance while meeting our customer's needs with cutting-edge, efficient, and budget-friendly solutions.

React Query empowers applications by efficiently managing server-state, while its minimalist API, requiring only two hooks and one utility, streamlines data fetching and manipulation tasks. This approach significantly reduces the need for manual data-handling logic, particularly beneficial when working with TypeScript, facilitating smoother development experiences. Discover the basics.

Panenco developed a customizable Retrieval-Augmented Generation (RAG) chatbot to address the challenge of managing vast e-learning content libraries and reducing repetitive support queries for a prominent client in the industry. By leveraging advanced reasoning and user-specific data, the chatbot provides personalized content recommendations, efficient query resolution, and engaging user interactions, demonstrating the efficacy of tailored AI solutions in enhancing content accessibility and user satisfaction.

In a recent cooperation we assisted a well-established SaaS company in the medical industry to renew their technical foundations. By migrating to Google Cloud Platform, updating technology stacks, and simplifying the API and database setup, significant improvements were made, ensuring a more secure, efficient, and maintainable system.

Advancements in AI services, exemplified by the development of Budd-E, an AI-powered remote control robot, demonstrate the integration of natural language processing capabilities into hardware devices, mirroring the transformative journey of the Internet of Things (IoT). Utilizing OpenAI's GPT-4 and LangChain framework, Budd-E interprets unstructured text commands into structured actions, showcasing the potential of AI to enhance the intuitive interaction between humans and technology.

A customer's cloud infrastructure, initially a self-managed Kubernetes setup on DigitalOcean, underwent a transformation to a fully managed solution on Google Cloud Platform (GCP), optimizing operations and scalability. By migrating components to GCP's managed services and implementing advanced monitoring and logging, the system achieved enhanced efficiency, reliability, and cost-effectiveness.

The Internet of Things (IoT) has transformed our interaction with the physical world, integrating digital sensing and control into everyday devices, from household appliances to city infrastructure. Stefanos Peros reflects on his PhD research which was focused on addressing time management challenges in IoT applications through innovative middleware and programming language concepts with new insights as software engineer at Panenco.

We developed a virtualized floor plan feature for our client to efficiently install and manage sensor devices in buildings, enabling real-time monitoring of environmental parameters. This enhancement streamlined data visualization, centralized analysis, and facilitated collaborative monitoring, while ensuring a seamless user experience through modern design principles.

The rise of public cloud and IoT has led to a surge in data generation, necessitating sophisticated analytics solutions. Grafana, an open-source platform, meets this demand by seamlessly integrating with diverse data sources, enabling comprehensive monitoring and analysis of both application performance and underlying infrastructure across various digital products.

Advancements in AI, particularly in machine learning and perception, have significantly improved robots' capabilities, fostering innovations in autonomous vehicles, robotic surgery, and service robots. The emergence of LLM services, exemplified by Budd-E, a Wi-Fi-controlled office service robot with AI-powered features such as Amazon Polly and GPT-4/Vision, showcases the potential breakthrough in human-machine interaction in various fields.

To improve the user experience, mailbox giants such as Outlook and Gmail allow the integration of applications that can extend the functionality of the mailbox. Gmail calls them ‘add-ons’ and Outlook calls them ‘add-ins’. Over the past few months, we've had the opportunity to work on both types of applications.

Given the complexity and potential for errors in manually deploying resources and managing microservice architectures across Azure, AWS, and GCP through their respective consoles, we turn to state-of-the-art Infrastructure as Code (IaC) tools like Terraform, which streamlines and automates the provisioning of infrastructure as we'll explore next.

Since the debut of ChatGPT, there has been a surge in enthusiasm to craft AI-driven applications or extend existing ones with AI capabilities. From chatbots and virtual assistants to documentation parsers and content generators, AI has reshaped the landscape of many industries.

We’ve been operating for over 9 years in the heart of Leuven, a region known for its high-tech research and development. In this article we’ll describe how we’re assisting innovative firms in the region with their digital product development activities and cloud engineering.

We're partnering with Launch.career to build new ways for students and young graduates to find a match with their dream employer.

We're glad to announce our partnership with Froomle to build a self-service portal to set up and monitor product recommendations.

We’re proud to announce our cooperation with DataCamp: a software engineering partnership with roots in the wonderful Belgian city of Leuven.

We're working shoulder-to-shoulder with the SoundTalks engineering team in building an animal health monitoring product, rooted in scientific research from KU Leuven.

Panenco and Sentigrate form a partnership to define a full-service product offering driven by sensors, data engineering, artificial intelligence and digital technologies.