Career

Infobip

Senior software engineer

After my master's, I returned to Anam Protect full-time, splitting my time between the core message-inspection service, bringing LLMs into production, and helping people around the department do the same. I'm the department's go-to engineer for text similarity and SMS fraud prevention, I help run large Anam Protect deployments around the world, and I mentor new team members.

These days I spend most of my time on a new internal application, an AI-assisted workspace for Infobip's partnerships with mobile network operators. More on that once it ships.

• 1000sdeveloper hours saved

  AI Knowledge Hub

  An AI assistant I built that answers questions about how our applications work, straight from their source code. Everyone from other engineers and product managers to solution and network engineers asks it instead of the developers, which has saved thousands of developer hours.

• 42.5k→850CPU cores

  Fuzzy LLM cache

  A text-similarity cache in front of our LLM classifier that handles 700 million requests a day. Near-identical messages reuse an earlier answer, so classification runs at sub-millisecond latency. Together with the text-similarity work I started in 2022, the improvements cut the compute needed from 42,500 CPU cores to 850.

• 45→90%traffic classified

  GenAI classifier for A2P traffic

  I championed a generative AI classifier for application-to-person messages. Doubling the share of traffic we can label made routing cheaper.

• <10 msat the 99.9th percentile

  New regex engine for packet inspection

  I led the migration of the service that applies rules to every network packet. Latency improved tenfold and CPU use dropped to a third.

• 100M+messages labelled

  LLM-powered sender branding

  This started as a hackathon prototype. It is now a distributed production system that labels hundreds of millions of messages at single-digit millisecond latency.

• 10improvements proposed

  Fraud leakage review for a major operator

  I defined a taxonomy for the fraud that gets through, analysed the traffic with our data science team, and proposed ten product changes.

• 180people in the department

  The go-to person for AI

  Engineers and non-engineers come to me about AI tooling and product integration. I brought data science onto the firewall products and taught a 35-person non-engineering team to use Claude Code.

Talks

• Classifying 700 million messages a day with quality comparable to LLMs, but without the LLM cost. A public talk at BipTalks, Infobip's engineering meetup in Zagreb, September 2026. Photos from the talk

Teaching assistant at FER

Student teaching assistant, alongside my master's

I helped students who got stuck on the coursework, showed them ways to solve the assigned tasks, and quizzed them on and graded their lab assignments.

Courses

• Machine Learning 1
• Advanced Algorithms and Data Structures
• Introduction to Artificial Intelligence
• Deep Learning 1

MSc in Computer Science

University of Zagreb, FER

Focused on deep learning, computer vision and generative AI. Graduated summa cum laude with a 5.0 GPA out of 5.0.

• Dean's Award 2022/23
• Dean's Award 2023/24

• Master's thesis, "Image Generation and Inpainting Using Diffusion Models". I trained a diffusion model that learns from images as the condition but takes text at inference time, then built a web app that edits photos of faces from written instructions, like an AI-powered Photoshop. Read the thesis (in Croatian)
• Won 1st place at the Algotrade hackathon in April 2024, out of 25 teams and 100 competitors. In under 24 hours, our team of four built a bot that trades resources and energy with other players' bots, tuned for maximum profit.
• Co-authored a paper, "Beyond Duplicates: Unleashing Transitivity in Quora Data Augmentation". If question A duplicates B and B duplicates C, then A duplicates C; we tested whether adding pairs like that helps classifiers. Read the paper
• Ranked 1st in 6 of 19 courses and in the top 3 in 11. My median rank was in the top 2.9% of each course.

Show all 19 courses with my rank

Score is my share of the course's points. Rank is my place among everyone who took the course that year.

Course	Score	Rank
Semester 1
Distributed Ledgers and Cryptocurrencies	90%	1st of 110
Web Architecture, Protocols, and Services	95%	1st of 64
Evolutionary Computing Optimization	95.5%	5th of 86
Machine Learning 1	93.7%	7th of 347
Advanced Algorithms and Data Structures	94.7%	9th of 376
Knowledge Management	90%	44th of 622
Semester 2
Parallel Programming	98.5%	1st of 269
Analysis of Massive Datasets	98.4%	1st of 76
Computational Methods in Modern Physics	100%	1st of 8
Deep Learning 1	98.5%	2nd of 185
Text Analysis and Retrieval	92%	12th of 52
Research of Public Opinion in Acceptance of New Technologies	89%	24th of 441
Semester 3
Competitive Programming	92%	1st of 39
Neural Networks	99%	2nd of 179
Computer Vision	94.6%	3rd of 80
Deep Learning 2	98%	3rd of 54
Distributed Big Data Processing	95%	3rd of 18
Data Visualization	91.9%	4th of 138
Intellectual Property in Industry	88%	151st of 466

Infobip

Student software engineer, part-time alongside my studies

My first job, split across two projects.

IoT: SIM card lifecycle

I helped redesign and rewrite the service that manages the lifecycle of SIM cards, breaking a monolith into microservices. Built with Java, Spring Boot, Kafka and RabbitMQ.

Anam Protect: SMS fraud

Anam Protect is Infobip's SMS firewall. It looks at every message crossing a mobile network and blocks the fraudulent ones in real time, for 1.2 billion subscribers worldwide. I worked on its production systems, which analyse millions of messages a day, building new features, fixing bugs, updating designs and writing tests.

• 4×faster cluster search

  Bitwise text similarity

  A new text-similarity algorithm built on bitwise operations. It sped up the search for clusters of near-identical messages and saved about 1,000 CPU cores.

BSc in Computer Science

University of Zagreb, Faculty of Electrical Engineering and Computing (FER)

Focused on software engineering and mathematics. Graduated with a 4.94 GPA out of 5.0.

• Dean's Award 2019/20
• Dean's Award 2020/21

• Bachelor's thesis on how to scale deep convolutional networks, comparing models on standard classification datasets for transfer learning.
• Ranked 1st in 6 of 32 courses and in the top 5% in 25. My median rank was in the top 1.4% of each course.

Show all 32 courses with my rank

Score is my share of the course's points. Rank is my place among everyone who took the course that year.

Course	Score	Rank
Semester 1
Mathematical Analysis 1	98%	1st of 733
Digital Logic	97.3%	4th of 742
Introduction to Programming	94.5%	7th of 721
Linear Algebra	98%	11th of 793
Communication Skills	92.6%	116th of 670
Semester 2
Physics	100%	1st of 810
Fundamentals of Electrical Engineering	95.8%	5th of 721
Mathematical Analysis 2	95%	9th of 818
Object Oriented Programming	88%	32nd of 836
Management in Engineering	86%	43rd of 364
Semester 3
Algorithms and Data Structures	99%	1st of 475
Probability and Statistics	98%	1st of 541
Computer Architecture 1	93.5%	5th of 484
Discrete Mathematics 1	98%	6th of 505
Information Theory	86%	16th of 705
Engineering Economics 1	90%	16th of 385
Semester 4
Operating Systems	98.5%	1st of 489
Databases	91.1%	5th of 511
Introduction to Theoretical Computer Science	93.2%	5th of 473
Web Software Development	93%	7th of 512
Communication Networks	94.5%	12th of 559
Risk Management	92%	393rd of 700
Semester 5
Statistical Data Analysis	93.5%	5th of 182
Selected Topics in Software Development 1	95.4%	6th of 101
Software Engineering	92.2%	9th of 446
Programming Language Translation	76.1%	22nd of 459
Environmental Sustainability and Climate Change Mitigation	85%	29th of 261
Semester 6
Design Patterns	93.7%	1st of 114
Stochastic Processes	96%	3rd of 40
Computer Security	96%	4th of 446
Basic Design Principles	95%	20th of 70
Introduction to Artificial Intelligence	89.9%	21st of 504

X. gimnazija "Ivan Supek"

High school, science and mathematics programme. Zagreb

• Valedictorian of my generation.
• Physics competition in my final year, the first and only competition I entered in high school. I shared first place in Zagreb and finished 7th in Croatia at the national round. Official results