Skip to content
View karaberke's full-sized avatar
5 followers · 9 following

Block or report karaberke

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. 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
karaberke/README.md

Hi, I'm Berke 👋

Computer science senior at UT Austin ('27) and Software Engineer Intern & Project Lead at Idaho National Laboratory. I build backend systems, LLM tooling, and RAG systems.

Featured work

Project What it is Stack
nuclear-data-aggregation-and-pipeline Platform I lead at INL for comparing nuclear cross-section data across 12 evaluated libraries. Bounded JVM subprocess concurrency, a memory-bounded TTL/LRU cache, and 297 tests. Python, FastAPI, Plotly Dash, NumPy, Docker
luminrag Source-grounded study app with dense/graph retrieval, source-linked citations, editable concept graphs, and browser-stored knowledge bases. Python, FastAPI, FastEmbed, NumPy, NetworkX, React
my-kitchen Self-hosted pantry and grocery app with exactly-once inventory updates and about 430 tests. TypeScript, SvelteKit, PostgreSQL, Docker
Comparative-RAG-Databases RAG app comparing Qdrant hybrid, ChromaDB, and Neo4j graph retrieval. Presented as a poster at INL. Python, FastAPI, React, Qdrant, Neo4j
fitness-tracker Mobile-first workout tracker PWA with per-user ownership enforced in SQL. TypeScript, SvelteKit, PostgreSQL
ag_news-comparative-study CS 371N team project comparing five text classifiers on AG News; fine-tuned DistilBERT reached 92.96% accuracy. Python, PyTorch, Hugging Face

Selected coursework projects

These are course projects. The linked repositories are cleaned portfolio copies of the submitted source; their READMEs describe supplied scaffolds and reproduction limits.

Project Work Tools
Energy-aware CPU scheduling Implemented a scheduling policy and compared 15 policies in a course-provided eight-core simulator. The selected policy used 83% less simulated energy than the provided round-robin baseline. C++
Kubernetes / Helm Completed Kubernetes assignments and submitted a Helm exercise using Minikube, kubectl, separate releases and namespaces, and a Flask endpoint with the Kubernetes Python client. Docker, Kubernetes, Helm, Flask
Database design / PostgreSQL / MongoDB Contributed to group database projects; the PostgreSQL project used AWS for coursework. SQL, MySQL, PostgreSQL, MongoDB
Motion autoencoder Submitted course source for motion generation and editing with a scaffolded PyTorch autoencoder; the README records incomplete paths and reproduction limits. Python, PyTorch
PintOS Group operating-systems project across threads, user programs, virtual memory, and file systems. Each submitted stage is preserved as a tag; the final file-system snapshot on main does not integrate every earlier implementation. C, Operating Systems

Tools

Languages: Python, C, C++, Java, TypeScript, SQL Backend & data: FastAPI, PostgreSQL, SQLite, Docker ML: PyTorch, Hugging Face, FastEmbed, ONNX Runtime, Qdrant

Contact

LinkedIn · berkekara@utexas.edu

Pinned Loading

  1. nuclear-data-aggregation-and-pipeline nuclear-data-aggregation-and-pipeline Public

    FastAPI + Plotly Dash platform for comparing nuclear cross-section data across 12 evaluated libraries via OECD-NEA JANIS. Bounded JVM concurrency, TTL/LRU cache, 297 tests.

    Python 2

  2. luminrag luminrag Public

    Educational question answering with source citations, editable knowledge graphs, Self-RAG verification, and browser-owned knowledge bases. Bring your own AI provider.

    Python 3

  3. nl-to-sql nl-to-sql Public

    Python

  4. ag_news-comparative-study ag_news-comparative-study Public

    Comparing five text classifiers on AG News: TF-IDF, Word2Vec + MLP, BiLSTM, DistilBERT (92.96% accuracy), and zero-shot BART. CS 371N team project.

    Jupyter Notebook 2

  5. Comparative-RAG-Databases Comparative-RAG-Databases Public

    RAG app comparing Qdrant hybrid retrieval (BGE-M3 + SPLADE), ChromaDB, and Neo4j graph retrieval behind one FastAPI API, with a React client and streamed LLM answers.

    Python 2

  6. my-kitchen my-kitchen Public

    Self-hosted Recipe and Kitchen Management App. SvelteKit, PostgreSQL, Drizzle, Docker.

    TypeScript