DISI Industry - Scholarships Program
Announcement for the granting of a number of scholarships for students enrolled in the last year of the study courses of the Department of Information Engineering and Computer Science, financed by companies on topics of corporate interest, including an opportunity to carry out an internship and thesis work within the company. For more details regarding the admission requirements it is advised to make reference to the official call available in the nearby download box.
The Scholarships listed below are those currently available for the summer call. Another call regarding Professional Training Scholarships will be issued around next October-November 2023.
Three typologies of scholarships are envisioned in DISI Industry initiative:
- Level 1 (L1) Scholarships: for last year Bachelor Degrees (students must be regularly enrolled at the second year at the moment of the call)
- Level 2 (L2) Scholarships: for last year Master Degrees (students must be regularly enrolled at the first year at the moment of the call)
- Professional Training (PT) Scholarships: for last semester of Master Degrees (students must be regularly enrolled at the second year at the moment of the call)
Before applying to a given scholarship it is advised to read carefully the corresponding skills and prerequisites (e. g., the degree for which it is proposed, technical skills..)
Call for academic year 2023-24 Level 1 Scholarships
APPLY NOW - deadline 8 September 2023 POSTPONED TO 30 SEPTEMBER 2023
Scholarship L1-Auxilium-GmbH
Scholarship Reference: L1-Auxilium
Funded by: Auxilium Cyber Security GmbH Siemensstraße 23, 76275 Ettlingen, Germany
Type of Scholarship: Level 1 - For students enrolled in the Computer Science Bachelor Degree Program
Title of Scholarship: Building-Up an internal OSINT Framework
Industrial Tutor: David Ernstberger David.ernstberger [at] auxiliumcybersec.com
Academic Supervisor: Bruno Crispo bruno.crispo [at] unitn.it
Short Description of Internship and Thesis Activities, and Expected Outcome:
About us: We are an independent information security consultancy company. We focus on providing high-value added service in multitude of security domains. Those include especially information security strategy design and implementation and penetration testing.
We have an exciting internship opportunity for a highly motivated and technically skilled student interested in open-source intelligence (OSINT). This internship will focus on planning and creating a framework designed to streamline the process of gathering public information about a specific target. The primary goal of this project is to create a comprehensive framework that guides users from a name, phone number, or email address to receive all relevant public information about the target.
Responsibilities:
Collaborate with the internal team to understand the framework requirements for the OSINT Process
Design and implement a flexible and scalable framework for automated OSINT data retrieval.
Implement data processing and filtering mechanisms to ensure the accuracy and quality of the retrieved information.
Conduct extensive testing to ensure the reliability and efficiency of the framework.
Document the OSINT process, including all Tools and techniques.
Benefits:
Gain hands-on experience in data retrieval, and OSINT techniques.
Work in a collaborative and supportive team environment.
Learn from experienced professionals in the field.
Enhance your problem-solving skills.
Build a portfolio showcasing your involvement in an innovative project.
Required Candidate Skills and Prerequisites:
Be a Bachelor student in Computer Science.
Familiarity with different OSINT Tools and techniques.
Knowledge of data processing and manipulation techniques.
Ability to work independently and collaboratively in a team environment.
Excellent problem-solving skills and attention to detail.
Good communication and documentation skills.
Call for academic year 2023-24 Level 2 Scholarships
APPLY NOW - deadline 8 September 2023 POSTPONED TO 30 SEPTEMBER 2023
Scholarship A-Adige Spa
Scholarship reference: A-Adige SPA
Funded by: ADIGE SPA – via Per Barco, 11 - 38056 - LEVICO TERME (TN)
Type of Scholarship: Level 2 - Reserved for students enrolled in Computer Science or Artificial Intelligence Systems Master Programs
Title of Scholarship: Three thematic options are made available:
- A1: Classification of generic 2D profiles shapes and definition of a computational solution for a feasible laser cutting sequence
- A2: Estimation of the manufacturing time of a generic 3D part on a laser tube machine, using a hybrid analytical-statistical approach
- A3: Automatic positioning of a free-form workpiece in 3D space for optimal laser processing in a robotized cell
Industrial Tutor: Paolo Benatti – paolo.benatti [at] blmgroup.it
Academic Supervisor: To Be Defined
Option A1: Classification of generic 2D profiles shapes and definition of a computational solution for a feasible laser cutting sequence
Short Description of Internship and Thesis Activities, and Expected Outcome:
The computation of an optimal laser cutting path for generic parts is currently completely automatic in case of metal tubes or profiles having a regular shape (round, square, rectangular, obround…). On the contrary, there is no automatic solution to compute a feasible and sub-optimal cutting path for profiles having special generic shapes, able to satisfy both the geometrical and the technological constraints.
In fact, there are many technical aspects which affect the definition of a reliable cutting trajectory in case of a generic “special” 2D section (interference between the part and the cutting head, flow of laser assistance gas on material surface, distance between the nozzle and the part, etc).
Scope of this Thesis is to study a large collection of existing, manually defined solutions, classify them and infer the underline rules for a feasible manufacturing, in order to compute an automatic trajectory path solution for special profiles.
The approach for the classification and the definition of the operational logic can be either classical, i.e. based on geometrical and topological analysis of the profile shape, or based on Artificial Intelligence solutions, e.g. Neural Networks of any kind.
The Expected Outcomes of this activity are:
- The classification of generic 2D shapes, stored in an existing database filled by expert technicians, clustering those that have similar cutting logic
This can be done using an AI solution or a more classical cluster analysis approach, like: o The identification of a finite number of numerical properties of a 2D generic shape, creating a correspondence between the shape and a point in a hyperspace o The definition of a metric in this hyperspace, where the distance between (the points corresponding to) two shapes expresses the “similarity” of these two shapes in term of geometrical and technological constraints (i.e. laser cutting paths have similar logic) - The Development of an algorithm, or an AI solution, able to find the group of shapes in the cluster which is “nearest” (more similar) to a given shape
This could lead to the development of a general algorithm, o AI solution, able to automatically compute a feasible cutting path sequence for any generic 2D shape.
Required Candidate Skills and Prerequisites:
The candidate should have acquired some theoretical knowledge of:
- Computational Geometry
- Cluster analysis, Data mining
- Machine learning, Artificial Intelligence, Neural Networks
Our technicians will provide training to reach a basic competence on laser cutting technology and related constraints
The candidate will work in cooperation with our Software Department, for the definition and implementation of the system and its underlying algorithms, and with our Application Engineers, for the analysis of the existing solutions and the tests of the outcomes on a group of new generic shapes.
Option A2: Estimation of the manufacturing time of a generic 3D part on a laser tube machine, using a hybrid analytical-statistical approach
Short Description of Internship and Thesis Activities, and Expected Outcome:
One of the most important information for users of automatic machines is the estimation of the manufacturing time of a given piece.
In fact, the processing time determines the production cost, which in turn allows a preliminary evaluation of the price at which (the production of) that piece can be offered to the market. A bad estimation of machining time (too high or too low) can lead to the loss of an order or, even worse, to a lost profit.
Unfortunately, for some technology (like lasertube machines), the estimation of production time with an analytical model is not an easy task, because there are many variables that influence the behaviour of the machine, affecting its efficiency: the creation of a digital-twin machine model is very complex.
A possible alternative solution to this problem is the exploitation of a statistical approach: given a large database of real data, collected from many machines on different parts with different materials and shapes, manufacturing time could be estimated by the extrapolation of the production time of “similar” parts, produced in “similar” condition (e.g: same machine model and capability).
Scope of this Thesis is to combine in a hybrid approach a preliminary geometrical analysis of a 3D part model, matching its geometrical features with a large database of real production data on a statistical base.
The candidate shall define some appropriate indicators to be automatically extracted by a geometrical analysis of a generic 3D part (like the shape of tube profile, the number of cutting geometries, the total cutting length, etc) and use these indicators to find the most similar parts produced in the past on the same material and obtain a range of min-max estimated production time, computed on the basis of their real production time.
The approach for the classification of the parts can be either classical, i.e. based on geometrical and topological analysis of the part itself, or based on Artificial Intelligence solutions, e.g. Neural Networks of any kind.
The Expected Outcomes of this activity are:
- The classification of the parts stored in an existing database filled by expert technicians, clustering those having “similar” features
This can be done using an AI solution or a more classical cluster analysis approach, like: o The definition of a finite number of indicators which identify the most important features of a given part (imported as a 3D model), creating a correspondence between the part and a point in a hyperspace
o The definition of a metric in this hyperspace, where the distance between (the points corresponding to) two parts expresses the “similarity” of these two parts in term of production time
- The Development of an algorithm, or an AI solution, able to find the group of parts in the cluster which is “nearest” (more similar) to any given part and the computation of an estimated production time range (min-max) for this part.
Required Candidate Skills and Prerequisites:
The candidate should have acquired some theoretical knowledge of:
- Computational Geometry
- Cluster analysis, Data mining
- Machine learning, Artificial Intelligence, Neural Networks
Our technicians will provide training to reach a basic competence on laser cutting technology and related constraints
The candidate will work in cooperation with our Software Department, for the definition and implementation of the system and its underlying algorithms, and with our Application Engineers, for the analysis of the existing solutions and the tests of the outcomes on a group of new generic parts.
Option A3: Automatic positioning of a free-form workpiece in 3D space for optimal laser processing in a robotized cell
Short Description of Internship and Thesis Activities, and Expected Outcome:
The manufacturing of a generic freeform workpiece using laser processes, like cutting and welding, is strongly dependent on the position and orientation of the part in the working space of the robotized cell.
In fact, to have a feasible laser process, each point along the cutting/welding path must be reachable by the laser tool with the correct orientation, respecting physical and technical constraints, like:
- Reachability by the tool (depending on the robot kinematics)
- Crossing of singularity points
- Interference among any of the objects in the cell
- Entanglement of cables and pipes
- Constant speed of the tool along its path (with respect to the technological process speed)
This is especially important for anthropomorphic robots, whose reachability space is not regular and (unfortunately) not very precise in absolute.
Scope of this Thesis is to map the space in the cell in term of reachability and precision and define a logic to match this space against the workpiece processing path, in order to compute the minimum number of positions/orientations of the part which allows the end effector of the robot to follow every point of the toolpath, respecting the aforementioned constraints.
The Expected Outcomes of this activity are:
- The preliminary analysis of the reachability space of a given robotized cell - The analysis of a certain number of real parts already processed by expert technicians, with the aim to generalize the logic behind their position and reposition in cell space. This can be done using an AI solution or a more classical geometrical approach. - The Development of an algorithm, or an AI solution, able to compute the minimum set of workpiece positions which allow full toolpath laser manufacturing
Required Candidate Skills and Prerequisites:
The candidate should have acquired some theoretical knowledge of:
- Computational Geometry
- Robotics and Kinematics
- Machine learning, Artificial Intelligence, Neural Networks
The candidate will work in cooperation with our Software Department, for the definition and implementation of the system and its underlying algorithms, and with our Application Engineers, for the analysis of the existing solutions and the tests of the outcomes on a group of new 3D workpieces.
The proposed topics have been summarized to fit the form provided by UNITN. Interested students will be able to better explore the proposals with the company managers involved, in order to satisfy training needs and expectations.
Scholarship B-Adige Spa
Scholarship reference: B-Adige SPA
Funded by: ADIGE SPA – via Per Barco, 11 - 38056 - LEVICO TERME (TN)
Type of Scholarship: Level 2 - Reserved for students enrolled in Computer Science or Artificial Intelligence Systems Master Programs
Title of Scholarship: Two thematic options are made available:
- B1: Evolution of a client-server architecture for remote diagnosis
- B2: Study and implementation of a 3D machine model
Industrial Tutor: Mattia Broilo – mattia.broilo [at] blmgroup.it
Academic Supervisor: To Be Defined
B1: Evolution of a client-server architecture for remote diagnosis
Short Description of Internship and Thesis Activities, and Expected Outcome:
Laser cutting machines are complex systems made by several moving units and multiple devices driven by a CNC. Each machine has an HMI PC Panel used to let the operator interact with the system. PC Panel collects information about what is happening from interaction with the operators, from motors, different sensors and devices with which the machine is equipped. On each machine there is also installed a server to let the interaction with the machine from remote, when machine is connected. This server exchange information over the network with several clients for different kind of applications: diagnosis, services operation, machine configuration, software update …
Scope of this thesis is to evolve the actual client - server architecture improving the communication channel and developing new useful features for remote machine and information management.
The Expected Outcomes of this activity are:
- Study of actual architecture of server installed on the laser cutting machines. - Study of actual situation of the remote connection of the machines
- Analysis and evaluation of the GRPC – dotnet library and the substitution in the actual architecture.
- Development and testing of the GRPC migration both on server and client side - Development of new functionalities in the architecture in order to improve communication with CNC, HMI software and the remote clients.
Required Candidate Skills and Prerequisites:
The candidate should have acquired some theoretical knowledge of:
- Computer science (programming skills)
- Computer Networks
- Telecommunication
- Human Machine Interaction
Our technicians will provide training to reach a basic competence on laser cutting machine, on the hardware and software architecture and related constraints.
The candidate will work in cooperation with our technical department, for the definition and study of the system and its underlying algorithms, in particular with our software engineers for the software implementation and testing.
B2: Study and implementation of a 3D machine model
Short Description of Internship and Thesis Activities, and Expected Outcome:
Laser cutting machines are complex systems made by several moving units and multiple devices driven by a CNC. It is important to have a panoramic view of the whole system and the real time state of all the moving parts. At the moment, our laser cutting machines have a 2D simulation component installed on the HMI PC panel on the machine that have this role. In order to improve the diagnostic process, and to evolve the existing framework for simulation and digitalization of the process, it is more and more necessary to evolve to a 3D solution.
Scope of this thesis is to evaluate the existing framework and toolkit available for developing a WPF component that can import/read simplified 3D model of machines and to use simulated or real time CNC input for moving objects and units.
The Expected Outcomes of this activity are:
- Study of actual architecture of the 2D simulation component installed on the HMI of the machines.
- Analysis and evaluation of 3D framework as Unity or Helix-toolkit.
- The development of component, able to load 3D models and configuration files that define devices and their role.
- Move components ad parts of the machine according to a simulated or CNC real-time input. - Make the component interactive in order to show or hide detailed information about the devices.
Required Candidate Skills and Prerequisites:
The candidate should have acquired some theoretical knowledge of:
- Computer science (programming skills)
- Robotics
- Human machine interaction
Our technicians will provide training to reach a basic competence on laser cutting machine, on the hardware and software architecture and related constraints.
The candidate will work in cooperation with our technical department, for the definition and implementation of the system and its underlying algorithms, and with our mechanical engineers, for the definition of the 3D simplified model.
The proposed topics have been summarized to fit the form provided by UNITN. Interested students will be able to better explore the proposals with the company managers involved, in order to satisfy training needs and expectations.
Scholarship C-Elixe SRL
Scholarship reference: C-Elixe SRL
Funded by: Elixe Srl - MPR, Trento, Via dell’Ora del Garda 99, 38121 Trento
Type of Scholarship: Level 2 - Reserved for students enrolled in Artificial Intelligence Systems Master Degree Program
Title of Scholarship: Control system for a magnetic robot for coating steel surfaces such as inside pipes or vertical surfaces of ships and tanks
Industrial Tutor: ALESSANDRO CONDINI - a.condini [at] elixe.com
Academic Supervisor: LUIGI PALOPOLI - luigi.palopoli [at] unitn.it
Short Description of Internship and Thesis Activities and Expected Outcome:
MPR srl and ELIXE srl are mechanical companies that study and manufacture machines for applying a polymeric coating for metals inside pipes and cylinders, operating in domestic and foreign markets.
Within the field of metal protection, robotic applications have developed mainly in surface cleaning and internal vision of pipes. Automation in the application of painting is still a little-studied field. Artificial intelligence still needs to be addressed.
In the realization of remote systems that apply protective paint, there are two areas in which MPR is interested in studying new solutions: 1) inside tubes even of small diameter (up to 40mm) and 2) on large vertical surfaces such as ships or tanks.
In both cases, it is necessary to define a system that analyzes the surface to determine the level of cleanliness and suitability for the application of the paint, then applies the paint, and finally verifies the quality of the protective coating.
The system must be autonomous based on existing magnetic robots. Still, it must be scaled and adapted to be battery-powered, equipped with a motor for moving, a vision system, and a tank connected to the paint dispenser. The process must be controlled to make the processing and the return to the charging and resting station.
The company can supply the mechanical part while the electronics and software architecture must be defined. Coordinating a mechanical study for the actuating device would also be interesting to optimize the weight and dimensions.
Required Candidate Skills and Prerequisites:
The candidate should be a student in the AIS Master, who is taking the intelligent robot specialization. He/she should have a good background in robotics from the previous education levels, or includes at least the following courses: 1. Introduction to robotics, 2. Robot Planning and its applications, 3. Optimisation and learning for robot control.
Moreover, the candidate will have to deal with the vision and control system of the machining operations together with the movement of the machine. A certain practicality is required to contribute to the execution of the electronics and the realization of the mechanics.
Scholarship D-Gruppo-GPI
Scholarship Reference: D-Gruppo-GPI
Funded by: Gruppo GPI Spa, Via Ragazzi del ’99, 13, 38123, Trento (Italy)
Type of Scholarship: Level 2 - Reserved for Master’s students in Computer Science, Artificial Intelligence Systems, Information and Communications Engineering
Title of Scholarship: Building Italian chat-bot for IT Help Desk by using transfer learning from ChatGPT and similar
Industrial Tutor: Paolo Ranzi - paolo.ranzi [at] gpi.it
Academic Supervisor: To be defined
Short Description of Internship and Thesis Activities, and Expected Outcome:
TITLE: Building Italian chat-bot for IT Help Desk by using transfer learning from ChatGPT and similar
DESCRIPTION:
The project aims to build a custom-made chat-bot. Such a chat-bot should leverage the Information Technology (IT) Help Desk’s data-set owned by GPI. The IT Help Desk is specialized in troubleshooting computer-related problems experienced by GPI's clients. Since the Help Desk is already running from several years by now, it has cumulated more or less 750000 records (called "tickets") about either common or less common computer-related problems. All tickets are written in Italian. Each ticket may be an email exchange between client and IT expert or a very short summary of the problem written by the IT technician.
METHOD:
The idea was to use an open source Large Language Models (LLMs) like LLaMA (Touvron et al. 2023) or GPT4All etc. as a backbone. By fine tuning/transfer learning the LLM will then be fed with data coming from the Italian data-set.
HURDLES TO OVERCOME:
- test whether the LLM generalizes to Italian data;
- although some tickets have a written step-by-step solution to the problem, some tickets are very succint (thus, no description of the steps undertaken for solving the problem);
- altough the interaction between client lamenting a problem and IT technician may happen as a telephone call, no calls were recorded. This means that a step-by-step solution taught by phone by the IT technican to the client is lost. Therefore we have to leverage the textual data only;
GOAL:
The ideal chat-bot should solve by itself at least 10 % of calls to the IT Help Desk. Indeed, we believe that at least 10 % of calls are simple and redundant IT problems (e.g. how to install new driver for the printer). If this works, this can save resources to GPI since some IT technicians of the Help Desk may work for other departments, instead of being busy at the Help Desk.
SCHEDULE (ROUGH ESTIMATE):
- month 1: read literature about anomimization and pseudo anonimization (see link below for an example); find and test optimal and GDPR-compliant software or procedure for anonimizing data-set;
- month 2: pre-processing data-set; find the right open source LLM;
- month 3: build the Deep Learning model (by transfer learning/fine-tuning) and train it;
- month 4: further iterations of the model and further training; test and validate the whole pipeline with unseen data;
- month 5: once the pipeline is stable, provide some help to developers with the deployment;
- month 6: writing thesis, documentation and polish code;
REFERENCES:
- https://www.enisa.europa.eu/publications/pseudonymisation-techniques-and...
- Touvron, H., Lavril, T., Izacard, G., Martinet, X., Lachaux, M. A., Lacroix, T., ... & Lample, G. LLaMA: open and efficient foundation language models, 2023.
Required Candidate Skills and Prerequisites:
- experience with Python programming (e.g. TensorFlow, PyTorch), specifically with an emphasis on Transformers and transfer learning;
- some knowledge of the scientific literature about Large Language Models (e.g. ChatGPT, GPT4All, LLaMA etc.);
- some understanding of Linux systems (for getting computational power) and version control (e.g. GitHub, GitLab etc.);
- solid understanding of statistics;
- good English skills.
Scholarship E-Gruppo-GPI
Scholarship Reference: E-Gruppo-GPI
Funded by: Gruppo GPI Spa, Via Ragazzi del ’99, 13, 38123, Trento (Italy)
Type of Scholarship: Level 2 - Reserved for Master’s students in Computer Science, Artificial Intelligence Systems, Information and Communications Engineering
Title of Scholarship: Creating version 2 regarding a Speech Emotion Recognition AI Agent
Industrial Tutor: Paolo Ranzi - paolo.ranzi [at] gpi.it
Academic Supervisor: To be defined
Short Description of Internship and Thesis Activities, and Expected Outcome:
TITLE: Creating version 2 regarding a Speech Emotion Recognition AI Agent
DESCRIPTION:
In GPI version 1 of a Speech Emotion Recognition AI agent is already running. In simple words, such and AI agent is able to detect 5 human emotions from human audio. The data-set consists in 9 Hrs of clean audio in Italian. The data-set is proprietary since it is owned by GPI. This AI agent is already implemented during the telehealth sessions offered by GPI. A telehealth session is a videocall between clinician and patient. The clincian can see on his screen in real-time those 5 human emotions in real-time. Every 4 secs an emotion is printed on the clinician screen. Whenever the patient stops talking, the AI agent stops itself automatically. The goal is to push such AI agent to a more advanced version (version 2).
METHOD:
Version 1 is based on the Deep Learning model developed in Patel et al. 2021, with the difference that a Variational AutoEncoder (VAE) has not been implemented. In a nutshell, version 1 takes an average of the
Mel-Frequency Cepstral Coefficient (MFCC) spectrogram. This is particularly useful since the deployment of the AI agent shows a 300 ms delay from the end of the 4 secs and the print of the emotion on the screen. Thanks to such an average, the AI agent performs almost real-time. The accuracy (computed against 30 % test-set) of version 1 is already around 84 %.
The challenge is to see whether the AI agent may be pushed towards version 2. In other words a VAE must be implemented. Further, the analysis of the whole spectrogram by using computer vision techniques and tranfer learning (see Luna-Jiménez et al. 2021) should be implemented, as well. These improvements should increase overall accuracy.
HURDLES TO OVERCOME:
- the version 2 (namely, the version computing the whole spectrogram) may be slower, thus attracting complaints from end users (i.e. clinicians);
GOAL:
Push the Speech Emotion Recognition AI Agent from version 1 (average of the spectrogram) towards version 2 (whole spectrogram).
SCHEDULE (ROUGH ESTIMATE):
- month 1-2: read literature about anomimization and pseudo anonimization (e.g. Qian et al. 2017); read relevant literature about Speech Emotion Recognition;
- month 3: pre-processing data-set; build the Deep Learning model (by transfer learning/fine-tuning) with VAE included; train it;
- month 4: further iterations of the model and further training; test and validate the whole pipeline with unseen data;
- month 5: once the pipeline is stable, provide some help to developers with the deployment;
- month 6: writing thesis, documentation and polish code;
REFERENCES:
- Luna-Jiménez, C., Griol, D., Callejas, Z., Kleinlein, R., Montero, J. M., & Fernández-Martínez, F.
(2021). Multimodal Emotion Recognition on RAVDESS Dataset Using Transfer Learning.
Sensors, 21(22), 7665.
- Patel, N., Patel, S., & Mankad, S. H. (2021). Impact of autoencoder based compact representation
on emotion detection from audio. Journal of Ambient Intelligence and Humanized
Computing, 1–19.
- Qian, J., Du, H., Hou, J., Chen, L., Jung, T., Li, X. Y., ... & Deng, Y. (2017). Voicemask: Anonymize and
sanitize voice input on mobile devices. arXiv preprint arXiv:1711.11460.
Required Candidate Skills and Prerequisites:
- experience with Python programming (e.g. TensorFlow, PyTorch), specifically Convolutional Neural Networks and VAEs;
- (optional) previous knowledge about scientific literature regarding Speech Emotion Recognition;
- some understanding of Linux systems (for getting computational power) and version control (e.g. GitHub, GitLab etc.);
- solid understanding of statistics;
- good English skills.
Scholarship F-Gruppo-GPI
Scholarship Reference: F-Gruppo-GPI
Funded by: Gruppo GPI Spa, Via Ragazzi del ’99, 13, 38123, Trento (Italy)
Type of Scholarship: Level 2 - Reserved for Master’s students in Computer Science, Artificial Intelligence Systems, Information and Communications Engineering
Title of Scholarship: Online model for robotics system status understanding
Industrial Tutor: Marco Lechtaler - marco.lechthaler [at] gpi.it
Academic Supervisor: To be defined
Short Description of Internship and Thesis Activities, and Expected Outcome:
The objective of this thesis project is to build an online model that can understand the health status of a robotic system. By analyzing the stream of input data, the model should be able to determine whether the robot is functioning properly or if there is a specific component that requires attention.
Objectives:
Develop an online model: design and implement an online model capable of analyzing a stream of input data from the robotic system in real-time and providing insights into its health status. The model should be able to classify the robot as either functioning properly or requiring attention/maintenance
Identify key features: identifying the crucial features or indicators that can be used to assess the health of the robotic system. These features may include sensor data, performance metrics, or other relevant parameters that provide valuable information about the system's condition
Find the trade-off between accuracy and performance on low-performance hardware. Optimize or adapt the model architecture, algorithmic optimizations, and efficient resource utilization to ensure the model can run effectively on resource-constrained systems
Activities:
Data flow establishment: establish a robust data flow system to collect relevant information from the robotic system for both training and production phases. This entails gathering data from sensors, performance logs, and other sources to construct a comprehensive dataset suitable for analysis
Key feature identification: analyze the collected data to identify the key features that correlate with the health status of the robotic system. This step involves statistical analysis, data exploration, and feature engineering to select the most informative indicators
Model design and implementation: design and develop the online model using artificial intelligence or machine learning techniques. The model should classify the robot’s components health status
Performance analysis: measure the accuracy and reliability of the model using the designated low-performance hardware.
Scholarship G-Gruppo-GPI
Scholarship Reference: G-Gruppo-GPI
Funded by: Gruppo GPI Spa, Via Ragazzi del ’99, 13, 38123, Trento (Italy)
Type of Scholarship: Level 2 - Reserved for Master’s students in Computer Science, Artificial Intelligence Systems, Information and Communications Engineering
Title of Scholarship: Miglioramento continuo dell'Application Lifecycle Management
Industrial Tutor: Zancarli David – david.zancarli [at] gpi.it
Academic Supervisor: To be defined
Short Description of Internship and Thesis Activities, and Expected Outcome:
Miglioramento dei processi di Application LifeCycle Management dalla compilazione all'analisi statica del codice fino all'analisi immagini docker dal punto di vista della sicurezza / vulnerabilità.
Attuazione e diffusione di nuovi standard metodologici e di strumenti.
L’attività dello studente consiste nello studio degli attuali processi di Application LifeCycle Management utilizzati in GPI e nell’identificazione di punti di miglioramento del processo in essere.
Sarà compito dello studente, identificare nuove soluzioni che consentono di migliorare il processo di Application LifeCycle Management:
- di compilare in modo automatico il software realizzato dalla Software Factory GPI
- di poter effettuare un’analisi statica del codice
- di effettuare lo scan delle vulnerabilità delle immagini docker prodotte.
Lo studente, avrà il compito di documentare il processo di Application LifeCycle Management al fine di poterlo diffondere velocemente a tutti i team di sviluppo del gruppo GPI.
Required Candidate Skills and Prerequisites:
Skill richieste / preferibili:
conoscenza di strumenti per CI / CD (jenkis / pipeline gitlab)
conoscenza base di docker
conoscenza basilare sugli scan per l’identificazione delle vulnerabilità software
conoscenza di strumenti per analisi statica del codice (es. Sonaqube)
conoscenza dei linguaggi java e degli strumenti per la compilazione del software
Scholarship H-Gruppo-GPI
Scholarship Reference: H-Gruppo-GPI
Funded by: Gruppo GPI Spa, Via Ragazzi del ’99, 13, 38123, Trento (Italy)
Type of Scholarship: Level 2 - Reserved for Master’s students in Computer Science, Artificial Intelligence Systems, Information and Communications Engineering
Title of Scholarship: ElasticSearch platform
Industrial Tutor: Matteo Stracchi
Academic Supervisor: To be defined
Short Description of Internship and Thesis Activities, and Expected Outcome:
The student will have to independently study the Elasticsearch platform to understand its functioning and potential with the aim of integrating these technologies into software solutions in the healthcare sector.
Elasticsearch is an open source search and analytics engine based on the Apache Lucene library. Elasticsearch is designed as a distributed Java solution to bring full-text search functionality in schema-free JSON documents across multiple database types. In particular, the project aims to create an application component that allows the quick search of content within personal, social and clinical data like a search engine similar to Google.
Required Candidate Skills and Prerequisites:
Basics of object-oriented programming in Java.
DISI Industry is the new service offered by the Department of Information Engineering and Computer Science
University of Trento
https://industry.disi.unitn.
www.disi.unitn.it