Hello! I'm Alper,

I'm currently working as a founding engineer at Freya, where we build our own TTS and STT models specifically for Turkish telephony. My background is in mathematics, thus I'm really interested in math-heavy part of the deep learning too. I like to practice interpretability and efficiency in LLMs and generative models. I'm particularly interested in transformer-based models, diffusion models, mechanistic interpretability and spectral analysis. Currently I'm interested in:

• Neural net pruning, specifically LLMs.
• Transformer-based models and architectural design strategies
• Training paradigms for LLMs, both in pre-training and post-training phases
• Offline/online self-improving agentic workflows
• STT, TTS, and full-duplex speech to speech models.
• Modeling the conversational dynamics of human speech.

PinnedLoading

1. Compact-Language-Models-via-Pruning-and-Knowledge-DistillationCompact-Language-Models-via-Pruning-and-Knowledge-DistillationPublic

   Unofficial implementation of https://arxiv.org/pdf/2407.14679

   Python 53 12

2. bookbotbookbotPublic

   A toy project for my generative AI studies on text data. Train generative models with given book/text files with just a single script.

   Python 1

3. fastapi/fastapifastapi/fastapiPublic

   FastAPI framework, high performance, easy to learn, fast to code, ready for production

   Python 95.4k 8.7k

4. huggingface/bloghuggingface/blogPublic

   Public repo for HF blog posts

   Jupyter Notebook 3.3k 982

5. huggingface/computer-vision-coursehuggingface/computer-vision-coursePublic

   This repo is the homebase of a community driven course on Computer Vision with Neural Networks. Feel free to join us on the Hugging Face discord: hf.co/join/discord

   Jupyter Notebook 784 235