Flagship case study
From an early AI product to a market-ready enterprise platform.
EagleGPT
When I joined BE BRAVE, EagleGPT had already been in development for approximately two months and an early functional frontend existed.
My role was to help take the product forward — shaping the user experience, translating customer needs into product requirements, coordinating development, testing and improving the product, and supporting its path to market.
- ~6 Months
- From early product stage to market-ready platform
- 3 First Customers
- Education · Food Manufacturing · Healthcare
- 7-Person Team
- Development collaboration
- 7 Open-Weight Models
- GLM · Qwen · Kimi · gpt-oss · Apertus
Product context
Enterprise AI without giving up data sovereignty.
EagleGPT is BE BRAVE's Swiss-hosted generative-AI platform for business use, designed around enterprise requirements such as data sovereignty, controlled infrastructure and integration into existing systems.
Swiss hosted
Business data is processed within Swiss infrastructure.
Company-specific AI
Company data can form part of the product's knowledge foundation alongside open-source LLMs.
Integration ready
API interfaces support integration into existing system landscapes.
Enterprise control
The product is positioned around security, compliance and data sovereignty.
Product characteristics as publicly described by BE BRAVE AG today — not a list of my personal contributions.
Why this mattered
Building the technology wasn't enough.
Customer problem
Companies want the productivity benefits of generative AI, but sensitive business information creates concerns around data location, privacy, control and integration.
Product challenge
How do you turn generative AI capability into a product companies can understand, trust, adopt and buy?
Product response — connecting:
- Useful AI capabilities
- Enterprise UX
- Swiss infrastructure
- Controlled data environment
- Integrations
- Customer-specific requirements
- Commercial positioning
My role
Connecting customer, product and execution.
01
Customer insight
02
Product requirement
03
Prioritization
04
UX & product decision
05
Development
06
QA & testing
07
Customer feedback
08
Go-to-market
My role was to translate what the product needed into actionable development work.
I created structured lists of features, changes and improvements for the external SELISE development team and worked closely with the team throughout implementation, testing and iteration.
Customer Discovery
Using customer conversations and feedback to understand problems, requirements and opportunities.
Requirements & Prioritization
Turning customer and business needs into clear product requirements — then deciding what to improve, change or build next, including UX and frontend structure.
Development Coordination
Translating product requirements into structured development and change requests for the external development team.
QA & Iteration
Testing implementations, identifying issues, providing feedback and reviewing improvements.
Go-to-Market & Commercialization
Contributing to positioning, pricing, launch activities and sales enablement — and presenting the product to customers and C-level stakeholders to support customer acquisition.
Product decision
One platform. Multiple AI models.
Different AI tasks do not always require the same model.
EagleGPT used multiple open-source AI models within one product experience.
When the existing setup did not deliver the desired experience for image generation, an additional model was introduced specifically for that use case.
Choose technology based on the problem — not the other way around.
Models integrated in EagleGPT
Different tasks used different models — selected for the job instead of fixed by default.
Z.ai
GLM-4.7
Reasoning & Tool Use
Qwen
Qwen2.5
General Purpose
Qwen
Qwen3-32B
Efficient Reasoning
OpenAI
gpt-oss-120b
Configurable Reasoning
Kimi
Kimi K2
Agentic Workflows
Kimi
Kimi K2 Thinking
Deep Reasoning
Apertus
Apertus
Swiss Open Model
Data sovereignty
Open-source AI. Hosted in Switzerland.
EagleGPT was built around open-source AI models running on Swiss-hosted infrastructure.
Customer data remained within the Swiss environment and was not used to train the underlying AI models.
Swiss Hosted
Infrastructure and data hosted in Switzerland.
Open-Source Models
AI models operated on Swiss infrastructure.
No Customer-Data Training
Customer data was not used to train the models.
From product to market
Product execution meets commercial execution.
The first three customers were won together with the CBO across three very different industries — showing the ability to translate the product into different business contexts.
Product technology
Swiss AI & Data Sovereignty.
During my time at BE BRAVE, EagleGPT already integrated Apertus, the open Swiss language model developed by ETH Zurich, EPFL and CSCS.
The integration fit EagleGPT's existing focus: Swiss hosting, data sovereignty and the secure use of AI within an enterprise environment, built on Swiss and European AI technology.
Learnings
What I took away.
01
Technology alone is not the product. The experience around the model decides whether customers understand and adopt it.
02
Choose the model for the job: image generation only delivered the desired experience once an additional model was introduced for that use case.
03
Trust is a feature. Swiss hosting and not training on customer data were product decisions, not footnotes.
Ready for the next step?
Let's see what we can build together.
Looking for someone who understands AI, moves products forward and thinks commercially?
Whether you're building a team or building a product, let's find out where I can create value.