Martin Ratajczak - Senior LLM Research Scientist

Recommended expert

Munich, Germany

Experience

May 2025 - Present

9 months

United States

Senior LLM Research Scientist

BYO Inc.

Research and develop models for chatbots, NLP and LLMs (e.g. Llama, Qwen, OpenAI)
Enhance chatbots with RAG, in-context learning
Supervised fine-tuning (PEFT, LoRA), Huggingface or Unsloth
Advanced training methods: Test-time training, (transductive) active learning, reinforcement learning
High-throughput serving with vLLM
Apply embedding models (e.g. SentenceTransformers), similarity/vector search or vector DB or ranking (e.g. LlamaIndex, Faiss, LangChain)
Generate and filter synthetic data, clustering
Detect hallucinations
Evaluate chatbot models (Rouge, BLEU, F1-Score, Recall, Precision)
Visualization of experiments (matplotlib)

Jan 2019 - May 2025

6 years 5 months

San Francisco, United States

Senior AI Research Scientist

Rev.com

Research and develop model architectures in speech recognition systems (ASR), large language models (LLM), natural language processing (NLP), speaker diarization, speaker recognition, text formatting, summarization and chatbots
Train and fine-tune neural networks and probabilistic models: CTC, Transducer, HMM, Segmental CRF, Conformer, Transformer, CNNs, RNNs
Train on multi-GPU nodes and large scale data sets
Optimize models for accuracy, size (e.g. quantization, pruning), and speed
Implement and optimize decoding algorithms
Data preparation: trained tokenizer, forced alignment, data cleanup scripts
Advise on road maps and quarter planning, write epics and tickets, supervise junior researcher and developer
Develop open-source ASR model as core member of research team
Publish first-author scientific publication at Interspeech 2025 on speech recognition and machine learning

Nov 2018 - Feb 2019

4 months

Munich, Germany

Machine Learning Engineer

e-bot7 - AI for Customer Service

Implemented and trained chatbots using neural networks and NLP methods

Mar 2017 - May 2018

1 year 3 months

Graz, Austria

Machine Learning Engineer

iTranslate

Implemented and trained speech recognition system (ASR) for mobile phones
Trained neural networks on multi-GPU system

Mar 2013 - Apr 2017

4 years 2 months

Graz, Austria

Research Project Assistant

Graz University of Technology

Researched neural networks and probabilistic models for sequences
Conducted machine learning (ML), speech recognition (ASR), and language model (LM) research
Innovated, implemented, trained and published papers on recurrent neural networks (RNNs), conditional random fields (CRFs), sum-product networks, new regularization methods and losses, calculation and coding of gradients for custom models, and segmental CRFs
Supervised master student project on machine learning and ASR
Analyzed Ca imaging recordings of neuronal activity of anesthetized and awake mice
Classified and visualized trajectories of brain states using dimension reduction (PCA, LDA), support vector machines (SVM), Kalman filter/smoother, clustering, regression, and time series forecasting

Sep 2008 - Dec 2011

3 years 4 months

Aachen, Germany

Research Project Assistant

RWTH Aachen University

Implemented and trained an end-to-end machine translation system based on conditional random fields (CRFs) from scratch including gradients and losses for a distributed multi-node CPU grid
Trained phrase-based statistical machine translation systems including language models (LM)
Trained and implemented log-linear models for text classification, part-of-speech tagging, named entity identification and syntactic parsing
Taught seminar and exercise courses in machine learning and pattern recognition including neural networks, statistical machine translation, and speech recognition (ASR)

Sep 2007 - Aug 2008

1 year

Munich, Germany

Software Engineer

GAF AG

Implemented software for geo-information systems, back-end server applications and web map services

Dec 2006 - Feb 2007

3 months

Munich, Germany

Graduate Research Assistant

Ludwig-Maximilians-Universität München

Analyzed and implemented online learning rules in neural networks

Feb 2004 - Aug 2004

7 months

Munich, Germany

Software Engineer Internship

Max Planck Institute for Physics

Implemented application to visualize learning processes (variants of Hebbian learning rule) and self-organization in neural networks (Hopfield network, associative memory, vector quantization, clustering, Boltzmann machine) for teaching
Implemented scripts using Fortran, Perl and C++ to visualize Large Hadron Collider experiments from CERN

Aug 2003 - Oct 2003

3 months

Munich, Germany

Software Engineer Internship

Siemens

Implemented a network of biologically-inspired spiking neurons

Summary

17 years of experience in machine learning (ML), 7.5 years of that in tech companies
Specialized in neural networks and probabilistic sequential models: automatic speech recognition (ASR), natural language processing (NLP), large language models (LLM) and generative AI, time series forecasting, classical machine learning and statistics, regression, classification, clustering, sequence-to-sequence models, supervised/unsupervised learning, anomaly detection, fraud detection
Solve tough algorithmic problems, design new architectures, build prototypes
Train and fine-tune models, optimize for accuracy, speed, and size
Integrate and deploy on cloud or on-premises (CPU, GPU, on-device, embedded)
Develop in Python (15y), PyTorch (6.5y), TensorFlow (4y), Java, C++
Research experience in machine learning at university labs and tech companies
Open to work as a freelancer, with a company, or as an employee
Publish scientific papers, speak at conferences, and present posters