Nikola Selic — AI Research Engineer

Finished My Thesis on Hybrid Retrieval

Mon, 10 Aug 2026 00:00:00 +0000

I’m happy to share that I have finished my thesis focused on hybrid retrieval.

PluraMath Preprint and Dataset Released

Tue, 07 Jul 2026 00:00:00 +0000

We released PluraMath, a human-curated benchmark extending mathematical reasoning evaluation to 18 underrepresented languages across six language families.

The work evaluates 27 reasoning LLMs and finds a persistent performance gap between high-resource and underrepresented languages. I contributed as a co-author and Serbian-language contributor.

Read the paper · Explore the code · View the dataset

Participated in the TUM.ai Makeathon

Sun, 19 Apr 2026 00:00:00 +0000

I had a great time at the TUM.ai Makeathon 2026 working on the Spherecast (YC S24) challenge.

Our team built Cubecast, a copilot for supply-chain disruption analysis in consumer packaged goods sourcing. It ingests a supplier delay email, extracts the disruption signal with an LLM, links it to supplier and product entities through a cognee knowledge graph and SQLite data, identifies the affected raw material, simulates the disruption, and suggests alternative suppliers and rerouting options.

We placed 6th out of approximately 45 teams in the Spherecast challenge.

Many thanks to TUM.ai for organizing the event, to Spherecast for the challenge, and to my teammates Demyan Kurbatov, Anton Komar, and Vranda Agarwal for an intense and rewarding weekend of building.

View Cubecast on GitHub

The Cubecast team

TUM.ai Makeathon 2026

Research Stay at CWI

Fri, 01 Aug 2025 00:00:00 +0000

From August to September 2025, I completed a research stay with the TRL Lab at Centrum Wiskunde & Informatica (CWI) in Amsterdam. I worked with Madelon Hulsebos and Daniel Gomm on LLMs, Text-to-SQL, and table retrieval.

Working at CWI

Centrum Wiskunde & Informatica

Amsterdam Science Park

Amsterdam

Attended ACM SIGMOD/PODS 2025 in Berlin

Fri, 27 Jun 2025 00:00:00 +0000

Had the opportunity to attend the 2025 ACM SIGMOD/PODS Conference in Berlin!

Started Internship at Firebolt

Thu, 03 Apr 2025 00:00:00 +0000

Excited to announce that I started an internship at Firebolt in Munich!

Joined Munich NLP as an Organizer

Tue, 01 Apr 2025 00:00:00 +0000

Excited to announce that I joined Munich NLP as an organizer, helping promote the field of Natural Language Processing and fostering exchange between students, professors, industry professionals and companies in Munich!

My Experience at BTW25 in Bamberg

Mon, 10 Mar 2025 00:00:00 +0000

I had such a pleasant time at the BTW conference at Fakultät Wirtschaftsinformatik und Angewandte Informatik (WIAI) in Bamberg this week. The organizers were wonderful, and kept me well caffeinated (thank you coffee bike). The lectures and workshops were such a treat for someone who is at the beginning of their research journey. Meeting so many people whose names I only saw on research papers was, as always, quite an experience. Many thanks to the organizers and the speakers. Special thanks to the TUMuchData team for the conference scholarship. Looking forward to the next one in Munich.

Why BTW25 stood out

BTW is an important meeting point for the German-speaking database community. The program offered a useful view of current database research while keeping the atmosphere open enough for students and early-career researchers to ask questions and meet the people behind the papers.

The poster sessions were especially valuable because they made it possible to discuss ideas directly with researchers and understand the practical decisions behind their work. Those conversations often provided context that is difficult to get from a paper alone.

Bamberg

The setting made the experience even better. Bamberg is a UNESCO World Heritage Site, and the old town provided a memorable backdrop for continuing conference conversations after the sessions ended.

Volunteered at CIDR 2025 in Amsterdam

Wed, 22 Jan 2025 00:00:00 +0000

Had the opportunity to volunteer at the Conference on Innovative Data Systems Research (CIDR) 2025 in Amsterdam!

Started new role as Teaching Assistant at TUM

Sun, 01 Dec 2024 00:00:00 +0000

Excited to announce that I started working as a Teaching Assistant for Introduction to Informatics (SQL and Java) at Technical University of Munich!

THINK.MAKE.START. Batch #18

Fri, 22 Mar 2024 00:00:00 +0000

I had a great experience taking part in THINK.MAKE.START. Batch #18, a two-week innovation program organized by TUM and UnternehmerTUM.

The program brought together interdisciplinary teams to develop and test new product ideas before presenting their prototypes on Demo Day. I met many smart and wonderful people working on innovative and fun projects throughout the two weeks.

Batch #18 concluded with Demo Day on March 22, 2024.

Many thanks to all the organizers and coaches who made the experience possible.

Demo Day presentations

The prototype fair

Started at Amazon Web Services

Thu, 01 Sep 2022 00:00:00 +0000

I’m happy to share that I started a new position as a System Developer Engineer at Amazon Web Services.

I joined the AWS EC2 KaOS team in Dresden, where I worked with machine learning and natural language processing for kernel security.

About

Mon, 01 Jan 0001 00:00:00 +0000

About

I completed my master’s studies at TUM, where I wrote my thesis on hybrid retrieval under the supervision of Jakob Sturm and Prof. Dr. Georg Groh.

My current focus is hybrid retrieval, LLMs, RAG, and retrieval over structured data. I am especially interested in how language models can find and reason over the right information.

Beyond information retrieval, my interests include information theory, measure theory, mathematical logic, and linguistics. I am always open to collaborating on interesting projects.

Before focusing on retrieval, I worked across data engineering, software systems, and security, including kernel-security work at AWS and LLM-assisted program analysis at Firebolt. That background shapes how I think about reliable AI systems today.

Research focus

Information Retrieval & RAG
Sparse, dense, and hybrid retrieval; retrieval evaluation.

LLMs & Structured Data
Table retrieval, Text-to-SQL, and reasoning over structured information.

Reliable AI Systems
Evaluation and systems aspects of deploying LLM-based applications.

My broader technical background includes databases, distributed systems, security, and machine learning for systems.

Education

Technical University of Munich

M.Sc. Computer Science · Enrolled 2021
Major in Machine Learning; minors in Databases and Security & Privacy
Master’s thesis on hybrid retrieval, completed August 2026

University of Novi Sad, Faculty of Technical Sciences

B.Eng. with Honours, Data Engineering
Thesis: Blockchain-based Federated Learning

Experience

Centrum Wiskunde & Informatica

Research Student, TRL Lab, 2025 · Amsterdam, Netherlands
Worked with Prof. Madelon Hulsebos and Daniel Gomm on generative AI and information retrieval for structured data
Completed a research stay from August to September 2025 focused on LLMs, Text-to-SQL, and table retrieval
TRL Lab people

Firebolt

Intern, April–August 2025 · Munich, Germany
Worked on a custom static-analysis workflow using large language models

Technical University of Munich

Teaching Assistant, December 2024–June 2025
Introduction to Informatics, with a focus on SQL and Java

Student Research Assistant, February 2024–December 2025

Twinu

Core Engineer, April 2023–January 2024 and February–September 2022 · Munich, Germany

Amazon Web Services

Software Engineer Intern, September 2022–February 2023 · Dresden, Germany
Applied machine learning and NLP to kernel security with the AWS EC2 KaOS team

KITRO

Software Developer, February–September 2022 · Remote
Worked with Python, Apache Spark, AWS Lambda, data engineering, and computer vision

LambdaWorks

Software Engineer Intern, February–May 2020 · Novi Sad, Serbia

Typhoon HIL

Machine Learning Engineer Intern, April–December 2019 · Novi Sad, Serbia

Community

Organizer, Munich NLP · April 2025–Present
Helping connect students, researchers, and industry practitioners in Munich

Volunteer, CIDR 2025 · January 2025
Conference on Innovative Data Systems Research, Amsterdam

You can also find my work on GitHub and Google Scholar.

Über mich

Mon, 01 Jan 0001 00:00:00 +0000

Über mich

Ich habe mein Masterstudium an der TUM abgeschlossen. Meine Masterarbeit über Hybrid Retrieval entstand unter der Betreuung von Jakob Sturm und Prof. Dr. Georg Groh.

Derzeit beschäftige ich mich vor allem mit Hybrid Retrieval, LLMs, RAG und Retrieval über strukturierte Daten. Besonders interessiert mich, wie Sprachmodelle relevante Informationen finden und darüber schlussfolgern können.

Über Information Retrieval hinaus interessieren mich Informationstheorie, Maßtheorie, mathematische Logik und Linguistik. Für die Zusammenarbeit an interessanten Projekten bin ich immer offen.

Bevor ich mich auf Retrieval konzentrierte, arbeitete ich in den Bereichen Data Engineering, Softwaresysteme und Security, darunter Kernel Security bei AWS und LLM-gestützte Programmanalyse bei Firebolt. Dieser Hintergrund prägt meine heutige Arbeit an zuverlässigen KI-Systemen.

Forschungsschwerpunkte

Information Retrieval & RAG
Sparse, Dense und Hybrid Retrieval sowie Retrieval-Evaluation.

LLMs & strukturierte Daten
Table Retrieval, Text-to-SQL und Schlussfolgern über strukturierte Informationen.

Zuverlässige KI-Systeme
Evaluation und Systemaspekte beim Einsatz LLM-basierter Anwendungen.

Mein weiterer technischer Hintergrund umfasst Datenbanken, verteilte Systeme, Security und Machine Learning for Systems.

Ausbildung

Technische Universität München

M.Sc. Informatik · 2021–2026
Schwerpunkt Machine Learning; Nebenfächer Datenbanken sowie Sicherheit & Datenschutz
Masterarbeit über Hybrid Retrieval, abgeschlossen im August 2026

Universität Novi Sad, Fakultät für Technische Wissenschaften

B.Eng. mit Auszeichnung, Data Engineering
Abschlussarbeit: Blockchain-basiertes Federated Learning

Berufserfahrung

Centrum Wiskunde & Informatica

Research Student, TRL Lab, 2025 · Amsterdam, Niederlande
Zusammenarbeit mit Prof. Madelon Hulsebos und Daniel Gomm zu generativer KI und Information Retrieval für strukturierte Daten
Forschungsaufenthalt von August bis September 2025 mit Schwerpunkt LLMs, Text-to-SQL und Table Retrieval
TRL Lab People

Firebolt

Praktikum, April–August 2025 · München
Entwicklung eines eigenen Workflows für statische Analyse mit Large Language Models

Technische Universität München

Tutor, Dezember 2024–Juni 2025
Einführung in die Informatik mit Schwerpunkt SQL und Java

Studentische Hilfskraft, Februar 2024–Dezember 2025

Twinu

Core Engineer, April 2023–Januar 2024 und Februar–September 2022 · München

Amazon Web Services

Software Engineer Intern, September 2022–Februar 2023 · Dresden
Anwendung von Machine Learning und NLP auf Kernel Security im AWS EC2 KaOS Team

KITRO

Software Developer, Februar–September 2022 · Remote
Arbeit mit Python, Apache Spark, AWS Lambda, Data Engineering und Computer Vision

LambdaWorks

Software Engineer Intern, Februar–Mai 2020 · Novi Sad, Serbien

Typhoon HIL

Machine Learning Engineer Intern, April–Dezember 2019 · Novi Sad, Serbien

Community

Organisator, Munich NLP · April 2025–heute
Vernetzung von Studierenden, Forschenden und Fachleuten aus der Industrie in München

Volunteer, CIDR 2025 · Januar 2025
Conference on Innovative Data Systems Research, Amsterdam

Meine Arbeit ist außerdem auf GitHub und Google Scholar zu finden.