Suggestions
Teppo Hudsson
Business-driven technology lead with strong international experience.
Teppo Hudsson is the Co-Founder and Chief Technology Officer (CTO) at Precode.ai, a company that specializes in AI design and automation solutions. He has played a pivotal role in establishing Precode.ai as a leading Nordic AI factory since its inception, focusing on providing cost-effective and rapid improvements to tasks through AI technologies.
Background and Education
Hudsson studied at London South Bank University, where he earned a degree in Graphic Design from 2003 to 2006. His educational background has likely contributed to his skills in design and technology, which are essential for his role at Precode.ai.
Professional Experience
At Precode.ai, Hudsson has been instrumental in developing innovative AI solutions, including bioprocess simulation models for the pharmaceutical industry. The company has been active since 2019 and has made significant strides in creating AI models that operate efficiently using scientist-owned data. Under his leadership, Precode.ai has also pioneered the AI gig economy, enabling freelancers and startups to build their AI products with ease.
Contributions and Achievements
- AI Development: Hudsson oversaw the creation of custom AI models that enhance customer service operations for companies like DNA Oyj.
- Innovative Pricing: He implemented a unique pricing structure based on data points generated rather than hourly consulting fees, making AI development more accessible.
- Community Engagement: Teppo Hudsson is also active on social media platforms like Twitter, where he shares insights related to AI and technology.
For more information about his professional background, you can view his LinkedIn profile under the username "hudsson" .1
Highlights
Are you sure you want to build some critical application on top of vanilla models?
I am betting that a testing and design elements will be critical soon in the world of AI. Follow up coming up on this.
To be clear, I am not expecting LLM models to have the knowledge necessary. They are designed for language generation, not in other ML tasks.
The image is generated 5min ago using vanilla ChatGPT model 4o.
Recently, a VP asked me about the failure rate of AI projects, and I estimated it to be around 25%.
There's also a lot of talk about AI agents, with people suggesting, "Just build the agents and connect them to make them smarter." Over the past five years of working on various AI projects, the main question that I would figure out is, how should you approach building AI agents when the failure rate is 25%?
Here's the reality:
- Be prepared for the project team to change direction four times while working towards the same goal.
- Each change will likely be complicated.
- Because of this complexity, the project will take more time and money.
Building AI agents often requires four times the effort, complexity, and cost to achieve a meaningful outcome.
If you'd like to know more or need help with AI design, feel free to reach out. Comment this post.
And I haven't even mentioned the effort needed to prevent silly questions, ensure security, check for bias or harmful content, and build reliable systems that can handle many users. Managing updates to documents, controlling costs, and maintaining response speed are additional challenges — but that's a whole other discussion.
