Skip to content
View Baltsat's full-sized avatar
🎯
Focusing
🎯
Focusing
12 followers · 11 following

Achievements

Achievement: Pull Sharkx3Achievement: QuickdrawAchievement: Pair ExtraordinaireAchievement: YOLO

Achievements

Achievement: Pull Sharkx3Achievement: QuickdrawAchievement: Pair ExtraordinaireAchievement: YOLO

Highlights

  • Pro

Block or report Baltsat

Report abuse

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

Report abuse
Baltsat/README.md

Konstantin Baltsat

I build the environments and graders that frontier models are trained and measured against — and I try to break them.

Merged open-source contributions

Small, sharp fixes in other people's codebases — mostly parser, formatter and SDK correctness bugs.

Project Contribution
highlight.js ★24k #4455 qualified generic type arguments in Java · #4453 where in Haskell GADT declarations · #4454 six-digit unicode ranges in CSS
pest ★4.8k #1184 restore stack state after a failed sequence
agntcy/dir #1892 register cloud KMS providers · #1891 avoid blocking on non-interactive key passwords · #1890 synchronize in-memory routing datastore
OpenFeature flagd (CNCF) #2006 synchronize readiness state
OpenFeature go-sdk-contrib (CNCF) #927 safely compare complex criteria values

Open and under review: undici #5601 (reject SharedArrayBuffer-backed body views), oxc #25022, pygments #3234.

Research

HL-EAI: A Multimodal Framework Enabling Emotional Reciprocity in Human–AI Strategic Decision-Making — Mozikov, Baltsat, et al., accepted at ACMM'25. Multi-agent behaviour in dictator-game, GTBench and trolley-dilemma settings, at AI Research Institute & ITMO LISA Lab.

MSc in Artificial Intelligence, ITMO University. BSc in Applied Computer Science, ITMO — Best Graduate 2024.

Remote contractor · Python, C++, PyTorch, JAX/MaxText, Terraform · konstantin.baltsat@gmail.com

Pinned Loading

  1. GraphRAG_benchmarking GraphRAG_benchmarking Public

    Benchmarking GraphRAG and AriGraph on MuSiQue, HotpotQA and BabiLong QA1, run locally with Ollama and Llama 3.

    Python 4 1

  2. Image-robustness Image-robustness Public

    Adversarial robustness of multimodal LLMs (CLIP, LLaVA): attack patches, Attack Success Rate, cross-implementation generalisation, and token-probability attack detection. Kaspersky-scholar research…

    Jupyter Notebook

  3. users-rosatom users-rosatom Public

    My Voice — clustering open-ended survey answers by meaning with multilingual BERT, UMAP/HDBSCAN and sentiment. Hackathon solution for Rosatom, F1 0.97.

    Jupyter Notebook 1

  4. Users-knowledge-search Users-knowledge-search Public

    Multi-platform document search over RAG, GraphRAG, NanoGraphRAG and LightRAG with a Telegram Web App. RAGAS answer relevancy 0.96. Digital Breakthrough international hackathon.

    Jupyter Notebook

  5. users-when-repair users-when-repair Public

    Healthy Pipes — predicting oil-and-gas pipeline failures from operating telemetry. Physics model plus ML, F1 0.91-0.99 on 3-fold CV. Hackathon solution.

    Python