arnesano/ jobs

DAVIDE ARNESANO / FOR STUDENTS

Software projects

Useful software for the student council. Build your own ideas as part of your degree.

Open opportunities

2 open opportunities
FaRaFIN · OVGU Magdeburg

Central FaRaFIN Calendar

Difficulty: easy
  • University software project

We are looking for students to deliver a central FaRaFIN calendar as part of their university software project, using an open-source solution or a self-hostable application. It should bring events together, support different groups and integrate with the existing infrastructure.

Responsibilities & profile

Calendar requirements

  • SSO and SMTP: integrate sign-in through the existing identity provider and email delivery through a configurable SMTP server. Agree on supported protocols and infrastructure together.
  • Group management and RBAC: manage dedicated calendars by group. Define administrator, editor and reader roles, and enforce access permissions on the server.
  • Personal view: show or hide authorised calendars according to personal preferences and save the selection. Hiding a calendar in the interface does not replace access control.
  • Calendar groups: provide separate, extensible areas for Events, Meetings, Veranstaltungen and Verteidigungen (thesis defences).
  • Notifications: provide subscriptions to selected calendars with email notifications to subscribed mailboxes, for example for new events, changes, cancellations and reminders. Include unsubscribe options and notification preferences.
  • Integration and compatibility: support .ics event downloads and calendar subscriptions via iCalendar feeds. An authorised read-only view should be embeddable in websites, for example using an iframe. Test display, updates and access controls with common calendar clients and browsers.

Quality and operation

  • Modern interface: clear desktop operation with accessible navigation and forms. Smartphone support is an optional enhancement, not a requirement.
  • Performance: fast calendar views and filters across multiple groups and larger event volumes. Agree on representative datasets and measurable response-time targets, and test them.
  • Privacy and security: process only necessary personal data and separate public from internal events. Address role checks, protected calendar feeds, TLS, deletion and retention rules. Notifications and embedded views must not expose private events to unauthorised users.

Approach

Start by evaluating existing calendars and suitable open-source or self-hosting solutions. Compare functionality, licensing, integrations and maintenance effort. Justify the selected solution and implement missing integrations or extensions. Building everything from scratch is not a goal in itself.

Outcome & collaboration

Deliver a working demonstration, documented selection decisions, configuration and any source code, tests for permissions and integrations, and instructions for deployment, updates, backups and handover to FaRaFIN.

Project support

Experienced students offer experience, ideas and advice. They do not contribute code; implementation is the responsibility of your project team.

To agree together

Scope, team size, technologies, schedule, SSO and SMTP access, and responsibility for ongoing operation will be agreed together at the start. Academic supervision and recognition must be clarified with the responsible teaching staff. Agree with your supervisor whether and how LLMs may be used.

How to get in touch

Briefly introduce yourselves, what you study and whether you are applying individually or as a team. post@farafin.de

Get in touch ↗
FaRaFIN · OVGU Magdeburg

Meeting Protocol Assistant

Difficulty: hard
  • University software project

We are looking for students to build an LLM-driven meeting companion as part of their university software project. The assistant listens to the meeting and transfers what was said into a preselected minutes template. It should work with different template types, Markdown first, and run on the local LLM cluster of the university.

Responsibilities & profile

Assistant requirements

  • Listen and transfer: record and process the audio of a running meeting, and map the content to the matching sections of the selected template, for example participants, agenda items, decisions and action items.
  • Generic templates: select a template before the meeting and add new templates without code changes. Support Markdown templates first; further formats should be possible later.
  • Local LLM cluster: run the models on the LLM cluster of the university, so that data stays in-house and costs stay low. We provide access to the server.
  • Chunked processing: process audio and text in chunks and merge intermediate results into one consistent set of minutes. Keep context across chunk boundaries.
  • Model choice: compare a pipeline of speech recognition and a text model with a native audio model. Justify the decision with measurements of quality, latency and resource use.
  • Review and edit: show the generated minutes before they are shared, allow corrections and export them as a Markdown file.

Quality and operation

  • Latency: results should be available during or shortly after the meeting. Agree on target values for delay and throughput on the cluster, and measure them.
  • Minutes quality: use representative test recordings to check that content arrives in the template completely and correctly. Make errors and hallucinations visible.
  • Privacy and security: process only necessary data, do not send recordings or transcripts to external services, and define deletion and retention rules. Participants must be informed about the recording and be able to consent.

Approach

Start by evaluating existing open-source tools for speech recognition and meeting minutes, and the models available on the cluster. https://github.com/TarunTomar122/better-voice is a useful starting point. Then build a lean prototype with one Markdown template and extend it step by step. Building everything from scratch is not a goal in itself.

Outcome & collaboration

Deliver a working demonstration on the cluster, documented model and architecture decisions, source code with tests for templates and chunked processing, measurements of latency and quality, and instructions for operation, updates and handover to FaRaFIN.

Project support

Experienced students offer experience, ideas and advice. They do not contribute code; implementation is the responsibility of your project team.

To agree together

Scope, team size, technologies, schedule, access to the LLM cluster, and responsibility for ongoing operation will be agreed together at the start. Academic supervision and recognition must be clarified with the responsible teaching staff. Agree with your supervisor whether and how LLMs may be used during development.

How to get in touch

Briefly introduce yourselves, what you study and whether you are applying individually or as a team. post@farafin.de

Get in touch ↗