Builder. Researcher.

Davyd Naveriani

BSc CS @ RWTH Aachen Β· ML Researcher Β· Exploring diffusion models for language, but open for all interesting ideas

Wrapping up my bachelor's degree in CS at RWTH Aachen while doing research in the Machine Learning and Human Language Technology group with Prof. Ralf Schlüter, Prof. Hermann Ney, and Dr. Albert Zeyer. Currently figuring out how discrete diffusion language models can help with speech recognition. Got a first-author paper under review, and I am looking forward to starting my master's degree with a lot more to learn and build.

DN

News


  • Apr 2026 πŸ‡¨πŸ‡­ Nominated for an exchange year at EPFL β€” super excited!
  • Apr 2026 πŸ₯‰ 3rd place at the HRT Trading Case at Datathon by Analytics Club at ETH
  • Apr 2026 πŸ“ˆ Selected for IMC's Accelerate Program in Amsterdam
  • Mar 2026 πŸ“„ Our paper on discrete diffusion LMs for ASR is out β€” fingers crossed for the review! arXiv:2604.14001
  • Jan 2026 πŸ“ˆ Selected for Optiver's Career Kickstarter in Amsterdam
  • Nov 2025 πŸ₯ˆ 2nd place at the IMC Trading Challenge at hackaTUM β€” great team effort!
  • Oct 2025 πŸ”¬ Joined the ML and HLT group at RWTH Aachen as a research assistant
  • Sep 2025 πŸ† Top 5 finalist at the Anthropic Hackathon (TUM.ai × CDTM, Munich)
  • Sep 2025 πŸ₯ˆ 2nd place at StartHack, AI Swiss Week in St. Gallen
  • Jul 2025 πŸ’Ό Started at Eggersmann Gruppe as a data science intern
  • Jun–Jul 2025 πŸ‡¨πŸ‡³ Summer school at HKUST (Guangzhou) β€” εΎˆζ£’ηš„η»εŽ†!
  • Feb–Jun 2025 πŸ‡°πŸ‡· Exchange semester at KAIST in South Korea β€” 정말 μ’‹μ•˜μ–΄μš”!

Research


Under Review

Diffusion Language Models for Speech Recognition

Davyd Naveriani, Albert Zeyer, Ralf Schlüter, Hermann Ney

Diffusion language models are becoming a real alternative to autoregressive ones β€” they can attend bidirectionally and generate text in parallel. In this work we look at how to actually use them for speech recognition. We put together a practical guide for rescoring ASR hypotheses with masked diffusion LMs (MDLM) and uniform-state diffusion models (USDM), and also propose a new joint-decoding approach that combines CTC acoustic information with USDM language knowledge at each step. Both USDM and MDLM noticeably improve recognition accuracy.

Education


Bachelor in Computer Science

RWTH Aachen University

Oct 2022 – Present Β· Aachen, Germany

  • GPA: 1.5 (German System), DAAD Scholarship Recipient
  • Dean's List β€” top 5% of cohort

Exchange Semester

KAIST (Korea Advanced Institute of Science and Technology)

Feb 2025 – Jun 2025 Β· Daejeon, South Korea

  • GPA: 98.5/100, Global Korea Scholarship (GKS) Recipient
  • Introduction to Deep Learning (A+)

Summer School

HKUST (Guangzhou)

Jun 2025 – Jul 2025 Β· Guangzhou, China

  • Intensive one-month program on mathematical foundations for AI
  • Probability theory & linear algebra

Work Experience


Student Research Assistant

Machine Learning and Human Language Technology Group, RWTH Aachen

Oct 2025 – Present Β· Aachen, Germany

  • Researching discrete diffusion language models for ASR rescoring
  • Implemented and trained a Diffusion Transformer from scratch at scale β€” see Research for details

Data Science Intern & Working Student

Eggersmann Gruppe

Jul 2025 – Present Β· Aachen, Germany

  • Built an anomaly-based predictive maintenance pipeline for 20+ sensor streams
  • Deployed model to production for real-time anomaly detection

Co-Founder & ML Engineer

orionic UG

Oct 2024 – Present Β· Aachen, Germany

  • Co-founded AI consulting firm, leading a team of 6 engineers
  • Built production systems: GitHub PR reviewer, RAG support bot, ETL pipelines

Student Research Assistant

ICoM, RWTH Aachen

Mar 2024 – Mar 2025 Β· Aachen, Germany

  • AI-powered data extraction for the KaSyTwin digital twin project
  • Built OCR extraction, database querying, and PDF parsing systems

Projects


Skin Cancer Detection with Swin & ViT

Experimented with two separate transformer models, Swin and ViT, to classify skin cancer from dermatoscopic images. A fun dive into medical imaging and vision transformers.

Python PyTorch Vision Transformers

CLIP for Plant Disease Detection

Took CLIP and adapted it for classifying plant diseases from leaf photos β€” compared fine-tuning vs. prompt-tuning to see which works better.

Python CLIP Prompt Tuning

RAG-Powered Discord Support Bot

Built this for an EdTech client at orionic. It handles ~70% of support tickets on its own and cut response times by 98%. Probably the most useful thing I've shipped.

LangChain FastAPI Docker AWS

Hackathons & Competitions


3rd place

HRT Trading Case β€” Datathon

Analytics Club at ETH Β· Apr 2026

Got 3rd on the HRT case with a Sharpe of 2.7. Our pipeline stacked Ridge regression, LSTM for temporal patterns, a delayed stream model, and sentiment features.

2nd place

IMC Trading Challenge β€” hackaTUM

Munich Β· Nov 2025

36 hours of building market-making and arbitrage strategies at hackaTUM. Placed 2nd out of ~40 teams (1,000+ participants overall).

Top 5 finalist

Anthropic Hackathon

TUM.ai × CDTM, Munich Β· Sep 2025

Built an AI voice agent for portfolio management β€” you talk to it, it gives you market insights and manages trades. Used Claude API and ElevenLabs.

2nd place

StartHack β€” AI Swiss Week

St. Gallen Β· Sep 2025

Made a financial dashboard that pulls in market news and auto-generates reports for clients. Placed 2nd in the case.

Blog Coming Soon


TBD

Diffusion Models That Write: A Practical Guide to Discrete Text Diffusion

A visual, intuitive walkthrough of how diffusion models work for text.