DataTalks.Club Meetups Global

Aleksandr Kim has spent nearly a decade doing what the industry now calls AI engineering, long before the title existed. As a Senior Data Scientist at Intuit in London, he builds AI-powered features in production at scale, and his career traces the full arc from data scientist to ML engineer and now AI engineer, where, as he puts it, the labels keep changing but the actual work barely has.

​In this conversation, Aleksandr unpacks one hard-won lesson from the front lines of enterprise AI: the model is rarely the win. He walks through the agentic system he architected at Intuit, one that aggregates data, auto-generates reports, and delivers them straight to leadership in Slack, saving over 30 hours of executive time a week, and explains why it only worked after he killed the original chatbot idea.
​
​What we get into:

• ​Why AI didn't shrink the gap between proof-of-concept and product. It just made the PoCs cheaper, and the unglamorous work between demo and deployment is still very much there.
• ​The agentic system that pivoted from chatbot to automation, and why customer interviews, not better modeling, were what made it ship.
• ​Translating ML metrics into business outcomes, turning precision and recall into things like First Contact Resolution and automation rate, on day one.
• ​Cost-efficient AI at scale, using cheap models for the easy cases and saving the LLM judge for the hard ones, cutting inference spend by about a third.
• ​Knowing when to abandon, the underrated senior skill of recognizing a problem that simply won't move, no matter the infrastructure or team.
• ​If you're an engineer or data scientist trying to figure out where your value sits in the LLM era, especially inside a large company, this one's for you.

​
​About the speaker:
​Senior Data Scientist at Intuit, based in London — doing what most people outside the company would now call AI engineering: building AI-powered features in production at scale. About 9 years across banking (Raiffeisen), cybersecurity (Kaspersky), retail (X5 Retail Group), and fintech (Intuit), with experience on both sides of the IC–management line: I led two data science teams at X5 Retail Group before moving back to IC at Intuit, and I founded and run Intuit's 70+ member Data Science Guild. Inventor on 15+ ML/AI patents.

​From SEC filings to market sentiment: Building intelligent financial analysis tools - Ivan Brigida

Agent demos often fail at the exact moment they become useful, due to a process crashes, interrupted deployments, or approvals that never trigger.

This workshop shows how to turn working LLM agents into durable production workflows. You will learn how to persist state outside the agent process, resume agents after failure, retry tool calls, pause safely for approvals, inspect every step of execution, and deploy real multi-agent apps to real infrastructure.

​We’ll cover the following steps:
​● Understand why agent demos break in production - Explore the common failure modes of LLM agents, including lost state, brittle long-running tasks, unreliable tool calls, unclear recovery paths, and lack of operational visibility.
​● Design agents as durable workflows - Learn how to structure agentic systems so each step is stateful, observable, retryable, and recoverable instead of being trapped inside a single fragile process.
​● Build a production-oriented agent pattern - We’ll walk through an example agent that uses planning, tool execution, decision points, and external services while preserving execution history and state across failures.
​● Add reliability controls for real-world execution - See how retries, timeouts, compensation logic, human approval, and event-driven continuation can make agents safer and more dependable.
​● Handle human-in-the-loop requirements - Learn where human review, approval, escalation, or correction should be inserted into agent workflows without breaking the overall execution.
​● Observe, debug, and improve agent behavior - Understand what needs to be visible when an agent runs in production, including execution paths, intermediate decisions, failed steps, tool responses, and recovery attempts.
​● Leave with a reusable production blueprint - By the end, participants will have a practical mental model for building durable agents that can be adapted to RAG systems, workflow automation, customer operations, data tasks, and enterprise AI applications.
​
​By the end of the workshop, attendees should be able to explain and implement the execution layer that separates a clever agent loop from a reliable agent service.
​● Recognize the production failure modes of in-process agent loops: process death, deploys, flaky tools, slow approvals, missing history, and distributed state.
​● Convert a basic tool-calling agent into a durable workflow with server-side state and per-step execution history.
​● Add a human approval step that can wait safely and resume without losing context.
​● Use execution traces to debug tool calls, LLM calls, timing, token usage, and failures.
​● Understand where durable agent execution fits alongside RAG, evaluation, monitoring, and capstone project expectations.
​

​About the Speaker:
Nicholas Lotz is a DevSecOps Engineer and technical enablement specialist dedicated to removing the organizational barriers that keep engineers from shipping great software. Currently a Technical Marketing Engineer at Voxel51 and a freelance DevSecOps consultant, Nick has built a career at the intersection of infrastructure automation and product education, including impactful roles at GitLab and Harness.

​He is the author of the second edition of Automating DevOps with GitLab Pipelines and is a recognized expert in Kubernetes, Terraform, and CI/CD modernization. With a unique academic interest in applying control theory to digital networks, he focuses on building transparent, secure software stacks that solve real-world business problems.
​