Skip to content
View nikolabakic's full-sized avatar
1 follower · 3 following

Highlights

  • Pro

Block or report nikolabakic

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
nikolabakic/README.md

Hi, I'm Nikola Bakić

4th-year Telecommunications & Signal Processing student at the University of Novi Sad.

I build evaluation-driven AI systems, with a current focus on RAG, information retrieval, LLM evaluation, and end-to-end ML applications.

Featured work

Project Focus
BankScope Evidence-grounded RAG, hybrid retrieval, LLM evaluation
Serbian LipNet Visual speech recognition, PyTorch, CTC
Wireless Autoencoder Deep learning for physical-layer communications
JND-LIC Learned image compression

Current focus

  • Retrieval-Augmented Generation
  • Information Retrieval
  • LLM and RAG Evaluation
  • Deep Learning and Multimodal ML
  • Reliable end-to-end ML systems

Core stack

Python · PyTorch · FastAPI · Qdrant · scikit-learn · Pydantic · pytest · SQL · React · TypeScript · CUDA

Connect

LinkedIn: linkedin.com/in/nikola-bakic

Pinned Loading

  1. Banking-Technology-and-Operational-Risk-Intelligence-Assistant Banking-Technology-and-Operational-Risk-Intelligence-Assistant Public

    Evidence-grounded RAG assistant for technology, cybersecurity and operational-risk research across U.S. bank 10-K filings. Hybrid retrieval, evaluation, FastAPI + React.

    Python 1

  2. serbian-lipnet serbian-lipnet Public

    Serbian visual speech recognition with LipNet (3D CNN + BiGRU + CTC). Speaker-disjoint evaluation with 41.20% test WER using beam search + a 5-gram LM.

    Jupyter Notebook 1

  3. misks-wireless-autoencoder misks-wireless-autoencoder Public

    End-to-end PyTorch autoencoder za bezicni prenos preko AWGN kanala (MiSKS, FTN UNS)

    MATLAB 1

  4. jnd-lic-portraits jnd-lic-portraits Public

    Implementacija ključnih komponenti JND-LIC metode (Pan et al., 2025) zasnovana na CompressAI baseline-u, sa evaluacijom na portretima iz CelebA-HQ

    Jupyter Notebook 1

  5. lip-language-classification-mu1 lip-language-classification-mu1 Public

    Visual speech (lip-reading) language classification — Serbian vs English — using HOG features, PCA, SVM, MLP and Random Forest. MU1 course project.

    Jupyter Notebook 1

  6. semantic-movie-recommender semantic-movie-recommender Public

    Transformer-based semantic movie recommender comparing MiniLM, MPNet and GTE ModernBERT embeddings on ~10K TMDB movies, with similarity evaluation, clustering and a Streamlit app.

    Jupyter Notebook 1