Developing scalable, ethical, high-performance AI
Details
Join our September meetup to learn about first-hand practical solutions and get expert guidance on building and managing highly effective AI systems and solutions.
LatticeFlow, Toloka, Swiss AI Association, Lightly ai, and Fartiq, all industry leaders, will discuss their experiences and the challenges they faced along the way.
This event will be of interest to CTOs, CPOs, Product Managers, and Heads of Products, Data Science, Analytics, and Marketing Leads, as well as other tech-industry business decision-makers.
19:30 - 19:55 – LatticeFlow
19:55 - 20:20 – Toloka ai
20:20 - 20:30 – Break for Networking
20:30 - 21:15 – Panel discussion with partners
Robust AI models with better data and model insights –
Pavol Bielik, CTO, LatticeFlow
Building high-quality AI models is a continuous process of training, validating, refining, and monitoring models in the presence of ever-changing data. This is known as the most time-consuming and expensive part of AI model development.
In this talk, we will demonstrate how to find and fix data issues in minutes — including wrong labels and inconsistent or missing annotations. We will also demonstrate how to plug-in custom state-of-the-art models (ranging from chest radiographs to preventive maintenance or sports analytics) and identify critical blind spots where the model systematically fails.
How we achieved superior scalability for processing social media mentions using Human-in-the-loop classification pipelines with adaptive ML models –
**Fedor Zhdanov, Head of ML, Toloka.ai**
Social media monitoring is an essential part of any social customer care strategy. Some marketing professionals disregard social listening due to its assumed complexity, but the reality is that if you don’t engage in social listening, you will be unable to help customers and will miss those who are interested in your products.
In this talk, we will discuss how to scale up SMM by building a human-in-the-loop pipeline for classifying mentions with minimal infrastructure effort. We will present a case study demonstrating how time to market can be significantly shortened by incorporating adaptive out-of-the box ML models into the classification pipeline.
Panel discussion –
Developing Reliable AI Solutions: Challenges and Best Practices
Moderator:
- Rok Zaloznik, Swiss AI Association
Panelists:
- Pavol Bielik, CTO, LatticeFlow
- Fedor Zhdanov, Head of ML, [Toloka.ai](https://toloka.ai/)
- Igor Susmelj, Co-Founder, Lightly AI
- David Wiedmer, Product Owner, Fartiq
Getting stakeholder approval of AI solutions, collecting and managing data, choosing the best models, providing continuous learning, and putting AI to use in the real world are just a few of the challenges that businesses face when adopting AI technologies.
We discuss these and other complexities in an effort to find a solution and provide expert guidance based on firsthand experience. In addition, we will talk about the forthcoming regulatory challenges and the steps that businesses and AI solution providers can take to effectively address them.