Dr. Oliver Borchers
Hello, and thanks for stopping by.
I'm Oliver Borchers. Drawing on 10 years of experience, I help managing directors and leadership teams implement AI initiatives well: which problem is worth solving, what gets developed and what does not.
For my doctorate at the University of Mannheim, I studied what deep learning can tell us about brands and customer experiences from product reviews and Twitter posts.
After that I moved from research into practice: first as a language processing engineer at Oxolo, then as CTO. Today I'm Head of AI Engineering at IU Group, responsible for the agentic AI systems in customer service. I also maintain open-source projects, including fse, which has more than 300,000 downloads.
Career
-
IU Group
Since June 2026
Head of AI Engineering
- Responsible for the AI agents in IU Group's customer service
- Technical lead for part of a team of around 15 people
- Service teams across twelve business units use our agentic customer service platform
- With the team, developed AI-supported workflows that make the first response to service requests about five times faster
- Requests more often reach the right team directly, with about 40% fewer handovers
- With the team, developed a reusable foundation for internal AI applications
July 2025 to June 2026
Staff AI Engineer
- Designed and built the agent platform behind the customer service automation
- Took it from first prototype to production
-
Floy
January to June 2025
Tech Lead Foundations
- Led the five-person Foundations team: data, DevOps, cloud and medical imaging
- Provided the platform and data pipelines for the AI teams, so they could build their own training datasets faster
- Increased the throughput of the radiology image data pipeline tenfold
- Migrated 25 services and five databases to AWS without downtime
- Brought Floy into the AWS Migration Acceleration Program (MAP)
- Carried out the migration in line with GDPR, SOC 2 and medical data requirements
-
Oxolo
January 2022 to December 2024
CTO
- First developed an AI companion for children with 3D avatars and a rendering pipeline, then a German GenAI video platform competing with Synthesia and Hour One
- Responsible for building the GenAI video SaaS platform and for the technical execution of the pivot to generative AI video
- The platform grew from zero to more than 600,000 users in 80 countries
- Led the technical due diligence for the €13M Series A with DN Capital and Headline and was responsible for board reporting
- Scaled the engineering and AI team to 25+ engineers; built hiring processes and engineering standards with the Head of AI and the Head of Engineering
- Ran an API with 99.99% availability at more than 5 million requests a month and halved AWS costs
- Responsible for SOC 2 readiness and GDPR compliance
November 2020 to January 2022
NLP Engineer to Lead NLP Engineer
- With the team, built the core NLP pipeline for Oxolo's original AI chat application
- From April 2021, led a team of three engineers and set development standards
-
University of Mannheim
January 2017 to November 2020
Doctoral researcher and postdoc
- Doctoral researcher at the Chair of Marketing & Innovation, January 2017 to April 2020
- Postdoc from April to November 2020
- PhD summa cum laude, awarded the Stiftung Marketing dissertation prize
-
UNC Kenan-Flagler Business School
October to December 2019
Research visit
- Research visit at the University of North Carolina's business school during my doctorate
-
University of Siegen
2011 to 2016
Bachelor's and master's studies
- B.Sc. Business Economics, 2011 to 2015
- M.Sc. Entrepreneurship and SME Management, 2015 to 2016
Doctorate
My dissertation is titled "Mining Unstructured Data for Market Insights: A Study of Representation Learning and its Value for Marketing". It examines how deep learning can extract market insights from unstructured data.
Representation learning means that a model learns on its own to represent texts and words as numerical vectors, which can then be used in further calculations.
-
Deep Autoencoder
Brands in product reviews
Deep autoencoders read Amazon product reviews and derive brand associations and competitive maps from them.
-
Word Embedding
How AI can forget
Dory is a temporal neural word embedding that forgets old associations and remembers new ones, so you can follow how brand associations change over the years.
-
BERT
Customer experience as a metric
The Customer Experience Score (CXS) is measured from Twitter posts using BERT and linked to financial metrics.
Open source
- fse GitHub stars: 626
fast sentence embeddings, more than 300,000 downloads. Available on PyPI; the fse organisation on Hugging Face provides 15 vector models.
- proqi GitHub stars: 45
A terminal tool for drafting the next instructions for coding agents while they are still working. Maintainer since August 2026.
- fractional-cto GitHub stars: 35
Public, MIT-licensed Claude Code marketplace.
- pydantypes GitHub stars: 23
Python library of validated Pydantic types for identifiers, ARNs, URIs and resource names in cloud, DevOps, data and AI code.
Publications
- 2019
fse: Fast Sentence Embeddings
Software
- 2019
Dynamics in the evolution of consumers' brand associations: A temporal neural word embedding for UGC
Marketing Dynamics Conference, College Park
- 2019
Dory: A temporal neural word embedding for the study of brand association evolution from user-generated content
41st ISMS Marketing Science Conference
- 2019
A model for temporal neural word embedding
48th EMAC Conference, Hamburg
- 2018
Automatisierte Marktanalysen von Produktbewertungen mithilfe Künstlicher Intelligenz
Report
- 2018
Deep neural language models: The elicitation of brand similarity from customer reviews
47th EMAC Conference, Glasgow
- 2018
Vector space models of brands
40th ISMS Marketing Science Conference, Philadelphia