20.500.12556/DKUM-98196 Razvoj platforme za nameščanje in upravljanje napovednih modelov strojnega učenja magistrsko delo Development of a platform for the deployment and management of predictive machine learning models Z naraščajočo uporabo strojnega učenja v praksi se povečuje tudi število napovednih modelov v produkciji, kar prinaša izzive pri njihovem učinkovitem serviranju. V magistrskem delu obravnavamo razvoj platforme za serviranje modelov, ki omogoča skalabilno, zanesljivo in učinkovito izvajanje napovedi. Poseben poudarek je na optimizaciji porabe virov, podpori sinhronemu in asinhronemu izvajanju ter uporabi interoperabilnega formata ONNX. Razvita rešitev je eksperimentalno primerjana z obstoječo platformo, pri čemer analiziramo odzivni čas, prepustnost in porabo virov. Rezultati kažejo, da lahko ustrezna arhitektura bistveno izboljša učinkovitost serviranja modelov. With the increasing adoption of machine learning in practice, the number of predictive models in production is also growing, which introduces challenges in their efficient serving. This master’s thesis addresses the development of a platform for model serving that enables scalable, reliable, and efficient execution of predictions. Particular emphasis is placed on resource optimization, support for both synchronous and asynchronous execution, and the use of the interoperable ONNX format. The developed solution is experimentally compared with an existing platform, analyzing latency, throughput, and resource consumption. The results show that an appropriate architecture can significantly improve the efficiency of model serving. serviranje modelov strojno učenje ONNX skalabilnost optimizacija virov. model serving machine learning ONNX scalability resource optimization true true false [M. Semprimožnik] Slovenski jezik Angleški jezik Magistrsko delo/naloga 2026-05-25 18:06:43 2026-06-29 12:22:36 2026-07-03 03:09:41 0000-00-00 00:00:00 2026 0 Maribor 0 Maribor 1 spletni vir (1 datoteka PDF (XI, 105 str.)) 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 004.85(043.2) 283408643 MAG_Semprimoznik_Matevz_2026.pdf MAG_Semprimoznik_Matevz_2026.pdf 1 1FC3858ED7E49E13C671E0B864D34EF5 cb86c6d910f2d62e486263e3d4e04f01f740b91fe5ff5323c386953492d569f8 691ea23e-6248-11f1-b32f-00155d580e34 https://dk.um.si/Dokument.php?lang=slv&id=206828 Fakulteta za elektrotehniko, računalništvo in informatiko 0 0 0