InnoCamp36
2nd Place with Team Fronius
Represented HAK Steyr at FH Steyr's InnoCamp36, working with Team Fronius on a challenge solution that earned second place.
About Me
I am Nikita Berger,
a Designer and Developer from Austria.
I am passionate about Artificial Intelligence, Mathematics, and Programming. I love working on side projects that challenge my problem-solving skills and deepen my understanding of these fields.
My goal is to specialize in Mathematics and AI, building innovative applications that improve lives and push the boundaries of technology.
Outside of work, I love doing sports, especially running.
Recognition
FH OÖ Diplomarbeitswettbewerb
Finalist, 2026
Reached the final with the BHAK Steyr Diplomarbeit team after presenting work on how AI can improve performance and efficiency.
JKU Aufgaben des Monats
Mathematics Finalist
Selected as a finalist in JKU Linz's mathematics challenge program for consistent, competition-level problem solving.
Cloudflight Coding Contest
Best Steyr Participant
Finished as the strongest participant from Steyr, placing 38th worldwide among 729 competitors in the Cloudflight Coding Contest.
Diplomarbeit
Wie Künstliche Intelligenz Leistungsfähigkeit und Effizienz steigert
A 2026 BHAK Steyr thesis on how artificial intelligence can increase performance and efficiency, created with Lara Sophia Harant, Felix Kühhas, Laura Gere, and Sandro Samuel Bramberger.
Selected for the FH OÖ Diplomarbeitswettbewerb and presented at Campus Steyr.
Small Language Model
Lille, a Compact and Powerful AI
Lille 130M is a 130-million-parameter language model designed as part of a fully open-source deep learning stack, offering strong performance in a compact size.
Available in two variants: Lille-130M-Base, pretrained on 4.27 billion tokens from the FineWeb-Edu dataset for general knowledge, and Lille-130M-Instruct, fine-tuned on the Kyoto-Corpus for instruction-based tasks like chat and command execution.
Powered by a modern Transformer architecture with Grouped-Query Attention, RoPE, and RMSNorm, Lille is optimized for efficiency and performance, making it ideal for research and deployment on modest hardware.
Created by Nikita Berger with ❤️
Knowledge Is Power
A User-Friendly Knowledge Platform
Knowpedia is a more user-friendly alternative to Wikipedia, designed to provide a seamless and engaging experience for users seeking knowledge.
It features a clean interface, intuitive navigation, and a focus on accessibility, making it easier for everyone to explore and learn from a vast array of topics.
Powered by Artificial Intelligence, it provides detailed answers on a wide range of topics.
Created by Nikita Berger with ❤️
AI Crypto Fund
Portfolio Value
The AI Crypto Fund is an innovative investment platform that leverages Artificial Intelligence to manage and optimize Cryptocurrency Portfolios. You pick the AI Agents and the Cryptocurrency you want to invest in, and the AI Crypto Fund will do the rest.
Chart data is for illustrative purposes only.
Created by Nikita Berger with ❤️
SophiaG Triton
High-Performance LLM Optimizer
Performance Benchmark (ms/step) — Lower is better
SophiaG Triton is my unofficial adaptation of the second-order optimizer by Liu et al., supercharged with Triton for superior GPU performance.
This project focuses on leveraging custom Triton kernels to accelerate computations, achieving up to a 3x speedup over the original implementation and outperforming standard optimizers like AdamW in mixed precision.
Created by Nikita Berger with ❤️