Among our global team of experts, Link Digital is lucky to have Tome Petrovski. Based in Bitola, a city in North Macedonia, Tome has high-level experience in the full Comprehensive Knowledge Archive Network (CKAN) tech stack, including Flask, PostgreSQL, Solr, and React for frontend improvements.

“Most of my work is focused on building and extending open data portals, like developing custom extensions, adding new features and maintaining existing functionality,” Tome says.

“I also spend time fixing bugs and improving overall system behaviour. It’s a role that lets me work on a variety of problems, which I really enjoy.”

Tome is now working on Link Digital’s ‘Ask AI’ feature, our new AI-powered layer on top of CKAN that seeks to make data portals more conversational, contextual, and semantically searchable.

I spoke to Tome about his work, his thoughts on making CKAN and the open data hosted on it more readable to AI agents that crawl portals and other specialised analytical tools, and how the 20 years he spent in the hospitality industry prior to joining Link Digital was perfect training for his current role as a software engineer.

Working behind the scenes on CKAN

In his time with the company, Tome has worked on many extensions. “Link Digital maintains quite a few custom CKAN extensions, and a big part of my work involves adding new features, fixing bugs and improving overall system behaviour.”

Some of these extensions include ckanext-charts, that enables users to create and manage interactive data visualisations directly within the platform, and ckanext-xloader, a CKAN extension that automatically loads data into the DataStore. But the work Tome is most proud of is his contribution to the Flask-Login integration, which also happens to be the most commented upon pull request in CKAN history.

The Flask-Login integration is the layer that allows users to authenticate in a portal, including managing user session login/logout and a ‘remember me’ functionality for Flask, a lightweight, open-source micro web framework for Python, designed to build web applications and APIs.

Before the upgrade, this function was handled by an authentication library that had been unmaintained for years. “It still worked but it had effectively become a black box within the community. No one fully understood how it behaved internally and that made it risky to modify or extend. No one was tackling bugs or adding new features.”

“So, I took on the task of replacing it with the Flask Login integration, which is a much better fit for the Flask architecture.” From a user’s point of view, nothing changed, but from a developer’s perspective, it replaced an unmaintained black box library with a modern, popular, and widely used extension that is a better fit for the CKAN flask-based architecture.

The main technical challenge of this eight-month-long project was ensuring backward compatibility with existing extensions that depended on implicit behaviour from the old system.

“A lot of existing extensions depended on implicit behaviour from the old system. So, we had to carefully replicate that behaviour to avoid breaking things. The work ended up being quite evolved. It ran for about 8 months with many review cycles, and it eventually became the most commented pull request in CKAN’’s history.”

Also read: Link Digital CKAN experts – global impact in 2025

Tome is one of the Link Digital CKAN developers now working on our new AI-powered layer for CKAN, Ask AI.

“The knowledge assistant is essentially an AI powered layer that allows a user to interact with open data in a much more natural way. Instead of manually searching through the data sets, users can ask questions and get direct, relevant answers. My work has focused on building the core engine behind it. How we retrieve the right information, generate accurate responses and ensure the system works reliably with the types of data typically found in CKAN portals.”

“One of the key priorities has been security. Since a lot of this data can be sensitive, we designed the system to support fully local deployment. This means the data never leaves the organisation’s infrastructure. While Ask AI also supports third party providers like OpenAI, the main goal is to give organisations the flexibility to keep everything in-house if needed.”

“I think open data portals have always had a discoverability challenge. You know the data is there, but finding and understanding it often requires some kind of technical knowledge. This kind of AI layer removes that barrier, making the data more accessible to a much wider audience. And doing that securely is what makes it truly viable for government and public sector use.”

The future of open data and AI

While Tome believes that the Ask AI feature is a major step in the right direction in terms of better aligning CKAN and open data management with advances in machine learning, through things like enabling smarter search and automatic metadata generation, it is “still relatively early stages.”

“One area I find particularly interesting is data quality. I think AI can make that much more intelligent, like going beyond validation rules to actually understanding the context of the data, then flagging inconsistencies that wouldn’t normally be caught and even suggesting improvements to publishers before data is submitted.”

“So, more broadly, I think there’s an opportunity to make systems more proactive… and guide publishers to produce higher quality data sets and reduce the manual effort involved on both sides. I believe that kind of intelligent layer could really raise the overall standard and make open data platforms more useful and accessible.”

Also read: What is the future of open data in the AI era?

The common data challenge that doesn’t get talked about

Tome’s work on Ask AI links to a much broader belief that discoverability is one of the core challenges facing organisations with their data portals.

“Organisations put a lot of effort into collecting and publishing data. But once the data is there, it often just sits unused. So, finding it and understanding it requires a level of guidance that most users don’t really have.”

Prior to Ask AI, “users were mostly limited to search filters or already knowing what they were looking for. Which means a lot of valuable data effectively stays hidden unless you’re familiar with the system.” This means that while their data is technically open, “in practice it’s not really accessible to everyone and I think that gap doesn’t get enough attention because the focus is usually on publishing the data rather than what happens after [it has been published].”

“That’s actually a big part of what motivated Ask AI. We are trying to bridge that gap and make data generally usable while still making it secure for government and public sector organisations to adopt.”

Lessons from working in the hospitality industry

Before transitioning into full-time software engineering, Tome worked for many years in the hospitality industry. It’s an experience he believes has been directly relevant to his job as a CKAN developer for Link Digital.

“I spent about 20 years in hospitality, starting when I was 18 and eventually managing cafes, restaurants and nightclubs. Honestly, that experience shaped me just as much as any technical training. Because running a venue is all about coordinating people, keeping the operation moving under pressure and staying organised when multiple things are happening at once. A lot of those skills translate directly into software engineering. Basically, you still need discipline, clear communication and the ability to stay calm and focused when something breaks at the worst possible time.”

“So, I think, yeah, coming into engineering with that kind of background was an advantage. I wasn’t just learning to code, I was already used to solving problems and delivering under real world constraints, which is ultimately what the job is.”

“The direction things are headed, the intersection of open data and AI feels like a genuinely important space, and I feel fortunate to be working in it.”

Interested in making your open data more machine-readable?

Get in contact to discuss how Ask AI can improve your users’ interactions with data.