Research & Publications

Academic contributions in Software Engineering, AI Systems, Cloud Computing, and Geodesy

Achievements

Competition wins, recognition, and academic distinctions

POPRI International Youth Business Model Competition — 1st Place

Representing Serbia with Sila | 2026

Won 1st place at the POPRI international youth business-model competition, representing Serbia with the Sila startup, and took part in a startup incubator. The win came with an invitation to Brussels for the EU Macro-Regional and Sea Basin Strategies forum.

1st Place Startup AI EU Forum

Innovation Competition — Photoristic

Science Technology Park Belgrade | November 2025

With team Photoristic — an AI-powered, real-time, embedding-only face-recognition solution for event photography — placed in the innovation competition at the Science Technology Park, validating the concept against established players in the space.

AI Face Recognition Computer Vision Startup

Mathematics & Academic Distinctions

National Competitions · MENSA Serbia

Long track record in national mathematics and physics competitions — roughly 15 mathematics and 11 physics awards during elementary school and 14 mathematics awards during high school, several at the state level. Member of MENSA Serbia since 2019.

Mathematics Physics MENSA State-Level Awards

Publications

Peer-reviewed papers in international journals and conferences

2025 Published

On the Applicability of Microservice Architecture in Geodetic Company

Novica Trifkovic, Ilija D. Antovic

2025 29th International Conference on Information Technology (IT), February 2025

This paper explores the practical application of microservice architecture patterns within geodetic companies, analyzing the benefits and challenges of transitioning from monolithic systems to distributed microservices.

2024 Published

Influence of Weights on the Unknowns' Estimation in 1D Geodetic Networks

Zarko Nestorovic, Novica Trifkovic

Journal of Faculty of Civil Engineering, Vol. 46, 2024

This paper presents research results on the influence of weights on unknown parameters estimation. The research was conducted for classical Least Square estimation and generalized matrix inversion across six cases of 1D networks, demonstrating that weight variations have relatively small influences on the unknowns within the considered range.

2026 Published

On the Statistical Base of 2D Shapes Similarity Determination

Novica Trifkovic et al.

ICIST 2026 — International Conference on Information Society and Technology

A statistical treatment of measurement uncertainty in 2D shape-similarity recognition using three-parameter and Helmert coordinate transformations, with error-propagation analysis and hypothesis testing on discretized shapes.

Forthcoming / In Review

Under Review

Towards Role-Based Multi-Agent LLM Systems for Software Requirements Analysis

Novica Trifkovic et al.

With FON (University of Belgrade) professors

Applies vertical, domain-specific AI agents to user-requirements gathering — a multi-agent, SLM-based approach (building on MARE and Elicitron) aimed at on-premise deployment for sensitive software requirements. It covers model selection and parameter-efficient fine-tuning (LoRA / QLoRA / PEFT), test-driven prompting, LLM-as-a-judge evaluation, and heterogeneous SLM↔LLM agent design, addressing the lack of empirical SLM studies for requirements engineering.

Multi-Agent LLM SLM / PEFT LLM-as-a-Judge Requirements Engineering
In Progress

Comparative Analysis of Cloud Providers across IaaS, CaaS, and FaaS (AWS, Azure, GCP, Hetzner)

Novica Trifkovic · PhD research

A holistic, reproducible benchmark of performance, cost, and developer experience across three deployment paradigms on four clouds, backed by a custom platform: one identical FastAPI app deployed to ten targets, measured by a neutral test client over a 14-day stratified campaign (N ≈ 28–31 per cell) and analyzed with non-parametric statistics. Empirical testing completed June 2026; write-up underway.