BEST PATH RESEARCH is an AI research, integration and development company.

Founded during the height of the COVID-19 pandemic, which saw the nature of many tasks, and how they are performed, change dramatically, BEST PATH RESEARCH seeks to capitalise on these once-in-a-generation changes to lifestyle and work, and bring its expertise in AI and Deep Learning to as many people and companies as possible.

Both our founders have been programming since their early teens and wrote and trained from scratch their first neural networks to recognize hand-written characters in their early twenties, when the first neural network boom was taking place during the early 1990s.

Since that time, they have both continued pursuing their academic and industrial interest in Machine Learning. Indeed, our founders first met at Philips Research Laboratories in Aachen, Germany in the early 2000s, where they were both conducting research into automatic speech recognition. Since the 2008 renaissance in (Deep) Artificial Neural Networks, they themselves have been deeply involved in both the theoretical and practical aspects of Deep Learning, which is now largely captured in the public imagination under the term “AI”.

Each member of our team has extensive experience using a variety of traditional Machine Learning technologies, as well as many of the recent (and not so recent) Deep Learning innovations such as CNNs, LSTMs, RNNs, and Transformers. They also have experience applying these technologies to an equally varied range of application domains, including handwriting and printed character recognition (OCR), object detection, named-entity extraction, speech recognition, and more.

BEST PATH RESEARCH therefore positions itself as a full-stack consultancy in all aspects of AI. We can advise you on all parts of your existing AI stack, or get you going if you are only just starting out. For example, we can handle everything from collecting and pre-processing the appropriate data for your task, to training and testing the models and deploying them in the cloud or on-device. We have proven deployments on iOS and Android with on-device AI inference, as well as cloud-based solutions that can be used interactively from a PC or smartphone, or in a “batch-mode” processing style for Big Data applications.

Get in touch today to find out how BEST PATH RESEARCH can help you make the most of the revolutions in AI.

Our Mission

Our mission is to research the best path to a solution for your problem using Machine Learning and AI. BEST PATH RESEARCH aims to make people’s lives easier and more fulfilling using AI technology, by assisting and augmenting people, not replacing them.

Our Vision

We envision a world in which machines assist people and businesses to get things done faster, more accurately and more efficiently, in particular by augmenting and complementing humans performing repetitive, manual and mundane work. We want to create a more efficient, sustainable and fulfilling life for everyone.

Our Team


Edward Whittaker

Dr. Edward Whittaker has over 25 years of experience in pattern recognition and machine learning.

He received his PhD in Statistical Language Modelling for Speech Recognition from the Cambridge University Engineering Department in 2000 where he was involved with the HTK broadcast news and conversational telephone speech transcription systems that consistently won the DARPA/NIST HUB4 and HUB5 evaluations in the late 90s.

After graduation Dr. Whittaker pursued post-doctoral research from 2000-2003 at the world-renowned Compaq Cambridge Research Laboratory (formerly DEC) in Cambridge, Massachusetts, and Philips Research Labs in Aachen, Germany. In 2003 Dr. Whittaker came to Japan to work at the Furui Laboratory in the Tokyo Institute of Technology. While there he continued his research into speech recognition and the development of a language-independent statistical approach to natural language question answering, a precursor to today’s purely data-driven machine learning approaches to the same problem.

Dr. Whittaker’s first company was formed in 2007 and was the first in Japan to develop a voice-activated train timetable search app in both English and Japanese for the recently-released iPhone. The app was sold to a major Japanese corporation and is still being actively used more than twelve years later. While helping develop various bespoke machine learning solutions for a number of clients in Japan he has also been an external developer and consultant at one of Japan’s largest internet companies.

Dr. Whittaker is the author, or co-author, on over 60 academic papers as well as several patents in the areas of compression and speech recognition. He currently also holds a part-time lecturing position at the University of Tokyo where he teaches a class on research and academic writing in the graduate school of Computer Science.

Selected Publications:
Ken-Ichi Iso, Edward Whittaker, Tadashi Emori and Junpei Miyake “Improvements in Japanese Voice Search” In Proceedings of Interspeech 2012
TREC 2005 Question Answering Experiments at Tokyo Institute of Technology E.W.D. Whittaker, P. Chatain, S. Furui and D. Klakow In Proceedings of the Fourteenth Text Retrieval Conference (TREC), 2005.
Efficient Construction of Long-range Language Models using Log-Linear Interpolation E.W.D. Whittaker and D. Klakow In Proceedings of 7th International Conference on Spoken Language Processing, ICSLP’02.
Quantization-based Language Model Compression E.W.D. Whittaker and B. Raj In Proceedings of the European Conference on Speech Communication and Technology, EUROSPEECH’01.
An Experimental Study of an Audio Indexing System for the Web B.T. Logan, P.J. Moreno, J-M. Van Thong and E.W.D. Whittaker In Proceedings of the 6th International Conference on Spoken Language Processing, ICSLP’00.
Comparison of part-of-speech and automatically derived category-based language models for speech recognition T.R. Niesler, E.W.D. Whittaker and P.C. Woodland In Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP’98
The 1997 HTK Broadcast news transcription system P.C. Woodland, T. Hain, S.E. Johnson, T.R. Niesler, A. Tuerk, E.W.D. Whittaker & S.J. Young In Proceedings of the 1998 DARPA Broadcast News Transcription and Understanding Workshop

Hans Dolfing

Dr. Hans Dolfing has over 30 years experience in the fields of software engineering, machine learning, artificial intelligence and biometrics. He graduated with a PhD thesis on the topic of handwriting recognition and verification and has worked in Europe and Silicon Valley on speech and handwriting recognition and verification, natural language processing, as well as signature verification, historical handwritten manuscripts, optical character recognition (OCR) and many other, surprising input signals. He contributes to industry and academic efforts, workshops and conferences from UNIPEN to JSALT. A distinguished industry career with many papers, patents and applications, from dedicated software solutions on tiny platforms, to applications for millions in daily life, in startups as well as large companies, visiting scientist at MIT, and experience with embedded and cloud-based AI.

Selected Publications:
Hans J.G.A. Dolfing, Whole page recognition of historical handwriting, arXiv:2009.10634, 2020.
Youssouf Chherawala, Hans J.G.A. Dolfing, Ryan S. Dixon, Jerome R. Bellegarda, Embedded Large-Scale Handwritten Chinese Character Recognition, arXiv:2004.06209, ICASSP 2020, Barcelona, Spain.
J.G.A. Dolfing, I.L. Hetherington, Incremental language models for speech recognition using finite state transducers. Workshop for automatic speech recognition and understanding (ASRU), Italy, December 2001.
J.G.A. Dolfing and A. Wendemuth. Combination of confidence measures in isolated word recognition. In International Conference on Spoken Language Processing (ICSLP), volume 7, pages 3237-3240, Sydney, December 1998.
J.G.A. Dolfing, J.J.G.M van Oosterhout, and E.H.L Aarts. On-line Signature Verification with Hidden Markov Models. In 14th International Conference on Pattern Recognition (ICPR), volume 2, pages 1309-1312, Brisbane, August 1998.

Get started with