praktykON

Agile teams vary in how fast they operate and how they estimate tasks. This poses challenges in objectively evaluating costs and efficiency across different teams.

Develop a universal cost estimation model for user stories that enables objective and measurable comparison of productivity across various Agile teams.

#Agile

The Weighted Shortest Job First (WSJF) method helps determine Agile priorities by accounting for both business value and the time required for implementation.
But does it actually increase team efficiency in practice?

Examine how WSJF affects the effectiveness of an Agile team. Analyse a case study to determine whether WSJF genuinely enhances prioritisation.

#Agile

As the number of scientific publications continues to rise, it becomes increasingly difficult to stay up to date with the latest research. How can millions of publications be automatically monitored and emerging trends identified before they are widely noticed?

Develop a solution for automatic monitoring of scientific literature that identifies emerging trends and research directions across various disciplines.

#Agile #Science

How can actual links between scientific publications and patent filings be identified and analysed for implementation potential in the absence of dedicated tools?

Develop a tool that automatically identifies links between research and patents within specific domains in order to anticipate directions of scientific and technological development.

#Innovation

Recent advances in technology and large language models (LLMs) have enabled natural conversations with AI driven systems.How can we design a proactive AI assistant with a visual avatar that not only communicates but also meaningfully supports users in health, education, and self development? 

Design a mobile AI assistant featuring a visual avatar that integrates an LLM with an animated character to offer perosnalised, in-depth support in specific areas. 

#AI

Experts in medical imaging can annotate the same lesion differently.These differences can significantly affect data quality and the performance of AI models.To what extent does the segmentation influence the stability of radiomic features and prediction accuracy? Can segmentation be standardized?

Analyse how region of interest (ROI) segmentation affects the stability of radiomic features and the quality of predictive models, and propose methods to improve repeatability and reliability of results.

#MedTech

In large IT systems, manual or simple rule-based task assignment quickly becomes ineffective. Some operators are overloaded, others remain idle. How can an algorithm ensure fair and scalable distribution of tickets, while accounting for availability, competencies, and actual workload?

Design a task scheduling module (Operator Scheduler) that automatically and fairly assigns tasks to operators, considering their availability, competencies, authorisations, and the ticket life cycle. 

#IT #Systems

Movement-based identification of individuals in VR works relieably as long as data is collected within a single session. When measurements are repeated after several days, model performance decreases. Why? Are performance issues due to trackers, their replacement or differences in movement between sessions?

Study key factors behind performance loss in cross-session VR identification and propose methods for mitigating the so called session effect.

* OPI PIB provides access to ready-to-use datasets, VR equipment, and software to record custom movement scenarios.  

#VR #MachineLearning #datascience

Can the way individuals move in a virtual environment reveal their emotions, personality traits, and psychological states? Immersive virtual reality enables precise control of the environment and detailed motion data collection. The key question is whether this data alone is sufficient to infer human psychology.

Develop tools that infer psychological states and personality traits exclusively from VR motion data.

#AR #VR #psychology

It is intuitive for people to recognize who is the leader, who dominates conversations, and who is less engaged. Can algorithms do the same? Group interactions appear complex, dynamic, and nuanced, particularly when analysed through audio and video data.

Design interpretable algorithms that analyse audio and video data to describe group roles, engagement levels, and interaction balance, without relying on a classic ‘black box’ approach.

#AI #socialinteractions

Electronic forms are widely used—from grant applications to registrations—but their data models fail to keep up with realities of modern collaborative workflows. Existing solutions do not provide flexible support for sharing, partial editing, and access control.

Design a technology agnostic data model and architecture for distributed forms that enables multi user collaboration; partial create, read, update, delete (CRUD) operations; and data consistency.

#IT

Data carriers and applications are still largely verified manually by security specialists. This approach is time-consuming and slows down ticket handling. This can be done faster, with higher quality and better scalability.

Design and implement a research workstation that automates the verification of data carriers and applications, integrated with ticketing systems, such as JIRA. Consider using AI for analysing licensing records.

#cybersecurity