My First Public ML Talk: Human-in-the-Loop Agent-in-the-Loop (Dec 16, DataTalks.Club)

#machinelearning #python #datascience #ai

About the Event

Modern ML systems are undergoing a significant shift: pipelines that once relied heavily on human-supervised feedback are gradually transitioning toward agent-driven, autonomous loops. In this session, we’ll walk through what actually changes—architecturally and operationally—when agents take on active roles inside ML workflows.

This will be my first public ML talk, and the focus is to deliver a clear, grounded, example-driven explanation of HITL → AITL transitions without unnecessary hype.

📅 Event Details

Date: 16 December 2025
Time: 12:30–13:30 CET
Registration: https://luma.com/d2o7pqt9

🔹 Outline

1. Human-in-the-Loop (HITL): Foundations and Limitations

What HITL means in practice (guided feedback loops, evaluation, corrections)
Key limitations: scalability, latency, reasoning gaps, context sensitivity
Where HITL still shines: accuracy, safety, human effectiveness

2. Agent-in-the-Loop (AITL): The Next Evolution Step

What changes when agents become active participants in the loop
Architectural view: planning, tool-use, continuous improvement, automation
Strengths: adaptivity, speed, scalable labeling, generalization

3. Real-World Comparison: When Each Paradigm Wins

HITL vs AITL across accuracy, trust, cost, transparency, and scalability
HITL-critical domains: medical, autonomous driving, manufacturing
AITL-favored areas: fraud detection, recommender systems, logistics

4. Transition & Future Outlook

Human-in-the-loop → human-on-the-loop → human-over-the-loop
Hybrid approaches combining human judgment with agent autonomy

🎤 Bio

Ertuğrul Mutlu is a Computer Engineering student at RWTH Aachen University and a Werkstudent Researcher at Fraunhofer IAIS (Enterprise Information Systems).

His interests lie in building reliable, lightweight AI systems, exploring agentic workflows, and combining classical methods with modern LLM-driven architectures. His work includes applied LLM engineering, evaluation pipelines, and signal-processing-inspired feature extraction.

He recently published a preprint on wavelet-based feature engineering and clustering, and writes technical articles on dev.to about ML systems, agentic design patterns, and practical AI engineering. This upcoming session will be his first public talk, marking the beginning of a broader effort to share practical insights and research explorations with the ML community.