Why This Session Matters
In this session, you’ll learn how to:
- Build low-latency AI voice agents for outbound operations and make architectural decisions that improve reliability, scalability, and compliance.
- Design evaluation, guardrails, and observability for agentic systems.
- Combine deterministic workflows with LLM-based intent understanding and simplify access to complex enterprise systems with AI assistants.
We are sharing a condensed, 25-minute version of the session, all the meat, no fluff, and general explanations.
What’s Inside
In this webinar, we cover:
Low-Latency Voice AI
(speech-to-text, latency masking, intent classification, voice UX)
Deterministic Agent Architectures
(conversation flows, structured outputs, tool execution, guardrails)
Enterprise Retrieval & Integrations
(Elasticsearch, chunking strategies, gRPC integrations, access control)
Evaluation, Security & Reliability
(LLM-as-a-judge, prompt injection testing, tenant isolation, reliability)
No sales pitches — just tested practices and frameworks.
Who Should Watch
This webinar is designed for:
- CTOs and Heads of Engineering
- AI and ML leaders
- Engineering managers
- Product leaders responsible for AI initiatives
- Teams building conversational AI, voice AI, or agentic systems
- Organizations evaluating AI automation for enterprise operations
Meet the Speakers
Mateusz Wosiński
Machine Learning Tech Lead, deepsense.ai
Mateusz is a Senior Lead Machine Learning Engineer specializing in building production-ready AI systems and leading agile, cross-functional teams. With deep expertise in Generative AI, computer vision, and MLOps, he focuses on transforming research prototypes into scalable, business-impacting solutions.
He has contributed to several high-impact projects, including:
- developing an application automating operations in a major HR platform using web agentic systems,
- building a semi-deterministic voice bot leveraging LLMs for intent understanding and context-aware responses,
- implementing an agentic system capable of autonomously resolving Jira and Git-based developer tasks using state-of-the-art LLMs.
Mateusz holds a Master’s degree in Applied Mathematics from the Lodz University of Technology and a postgraduate degree in Deep Learning from the Warsaw University of Technology.
