AI that shows its work
vinTsys builds AI systems that cite their sources, verify their results, and earn the trust of researchers and institutions.
See it in action with ScholarInt and our Learning Hub.
ScholarInt — an AI learning assistant that grounds every answer in your institution's own material.
Learning Hub — the technical journal, written from the inside.
ScholarInt — AI Learning Infrastructure for Institutions
ScholarInt is a guided learning platform that provides an intelligent layer for academic workshops and courses. It pairs a structured, stage-based curriculum with a Q&A assistant grounded in your workshop's own course material — so every answer is source-backed, with traceable citations, as you work through the content.
Currently deployed in bioinformatics and computational life sciences education, ScholarInt runs on workshop materials from the BioNT European training network. The platform is designed to extend to any domain where reproducible, citation-backed learning matters.
- Grounded answers, not guesses — every question you ask is answered by retrieving relevant passages from the workshop's course material (slides, papers, documentation) and generating a response from those passages, with sources cited.
- Stage-based curriculum — content is organized into sequential stages, each with its own learning objectives, topics, and a short quiz to confirm understanding before you advance.
- Progress that's tracked for you — your stage progress and quiz attempts are saved to your account, so you can pick up where you left off.
- GDPR-compliant by design — data stays within a private cloud environment
ScholarInt is in final development. We are onboarding a small number of institutional partners ahead of public release.
LLM Architecture & Agentic Systems — Learning Hub
Pubudu's public learning journal documents the internals of large language models and agentic systems — from tokenisation and attention mechanisms to context engineering and trust. Written from the perspective of a genomics researcher who built these systems from the ground up, not from the outside in.
Read at learn.vintsys.com →Building Blocks of GPT-2
tokenisation, embeddings, attention, transformer architecture
Architecting Autonomous Systems
agent anatomy, MCP, context engineering, and engineering trust
Recognized in BioNT's Sustainability Plan
BioNT is an EU-funded consortium developing open training resources in bioinformatics and computational life sciences. ScholarInt is built as an intelligent layer on top of these openly available materials, providing the structured learning experience, adaptive assistance, and assessment infrastructure that institutions need to deploy them at scale.
vinTsys's work is documented in two of BioNT's official deliverables to the European Commission: the project's Third Year Impact Assessment, which records ScholarInt as a working example of the consortium's materials in active use, and BioNT's Sustainability Plan, where vinTsys — registered as a Norwegian spin-off — is cited as a concrete case of how BioNT's training resources can be reused and sustained beyond the project's own funding.
In effect, an EU consortium's own reporting to the European Commission holds up vinTsys as evidence that its work outlives the project.
What Else We Can Build
ScholarInt is our primary product. The work below demonstrates the broader platform capabilities vinTsys can bring to learning and knowledge systems.
Multi-Turn Conversational Learning Agent
A conversational agent that guides users through real-world language practice scenarios — cafés, travel, everyday situations. It tracks vocabulary mastery across sessions, identifies weak areas, and adapts its approach based on what the learner gets wrong.
Built as a platform capability demonstrator: persistent user profiles, multi-turn memory, and MCP integration — agentic architecture beyond RAG pipelines.
Built because the problem was too important to ignore
AI is reshaping research and education. However, current generic AI tools share a fatal flaw: they produce answers with citations that cannot be verified against your own domain-specific materials and operate on generic knowledge rather than the specialised content your work actually depends on. This can leave institutions with no way to audit what they're getting. In high-stakes research and education, that is not a gap — it is a liability.
The mission of vinTsys is to build AI systems grounded in your domain's own content. We aren't here to ride the AI wave; we are here to build traceable, verifiable systems that institutions can actually trust.
vinTsys is currently focused on AI in education and professional training — building systems that institutions can deploy with confidence. That focus is deliberate: education is where trust in AI is most visibly broken and where the impact of getting it right is immediate. The same foundations that make this possible — domain-grounded content, traceable and verifiable outputs — extend naturally into active research. Our roadmap moves in that direction: giving research teams customisable, verifiable AI that operates on their own materials, in their own domain, on their own terms.
Built on a career of rigor
vinTsys was founded by Pubudu Samarakoon. Across a diverse career spanning a PhD in Bioinformatics at the University of Oslo, postdoctoral research in cancer genomics at Oslo University Hospital, and industry ML software development for personalized immunotherapy, Pubudu developed a deep reliance on domain-specific rigor and verifiable data.
When generic AI tools began transforming the tech landscape, he immediately recognized their fundamental limitation: they lacked the domain-specificity and traceability required for high-stakes environments. Whether building complex pipelines for computational biology or delivering ML and LLM training as a Senior Engineer at the University of Oslo, he saw firsthand how unverifiable outputs can affect advanced workflows. vinTsys is his response: providing tools that can adapt to complex domains—starting with education, and scaling directly into advanced research and computational biology.
Interested in ScholarInt for your institution?
We are working with a small number of training networks, bioinformatics core facilities, and PhD programmes ahead of public release. If you run workshops or structured training in computational life sciences, we'd like to hear from you.
Or write directly: hello@vintsys.com