How I Talk About LLMs in Tech Interviews (And The Systems I Built to Prove It)

Over the past year of navigating high-level platform engineering loops, I’ve found myself having the same deep-dive conversations: how to translate vague business questions into robust, event-driven AI pipelines. This video is my way of open-sourcing those conversations. I’m sitting down to unpack the exact nuances, jargon, and architectural decisions that come up when discussing production-level LLM systems. Instead of just talking theory, I’m walking through my actual portfolio—including an agentic pipeline built on a cascade of models and event-driven microservices. Timestamped Key Takeaways: 00:01 - Introduction to the interview presentation format 02:47 - Explaining the 'SWE Compass' mental model for system design 10:06 - Philosophy of open-source, zero-cost architecture 16:09 - The 'Show, Don't Tell' strategy for building interview credibility 20:46 - Defining Agents vs. Workflows in technical interviews 30:40 - Deep dive into LinkedIn data analytics and engagement optimization 41:58 - Architectural breakdown of the job board data pipeline 47:06 - Demonstrating the agentic writer and resume tailoring tools 56:45 - Using GitHub Actions logs for observability in production-level systems 01:00:48 - Managing asynchronous scheduling and token limits in AI workflows Main Takeaways / Conclusion: Effective Communication: When presenting projects, ground your technical explanation in the business background, the specific problem solved, and the architectural trade-offs to signal seniority. Prioritize Data Observability: Move beyond just showing code; demonstrate your ability to monitor, debug, and optimize systems using logs and performance metrics to show true production experience. Build 'Real' Projects: Focus on solving actual professional problems (like job hunting or content distribution) rather than generic tutorials, as this provides tangible artifacts to discuss during interviews. Own the Jargon: Be prepared to define and defend your architectural choices (e.g., RAG vs. Cache Augmented Generation) and embrace constructive feedback from experienced interviewers to iterate on your systems.