AI systems · LinkedIn
What Building AI Demands in 2026: Lessons from Shipping, Breaking, and Fixing Real Systems
Why AI products need stack ownership, bounded trust, continuous evaluation, and judgment formed in production.
Back to profile
Publication index
Writing
A curated record of published work. New writing will eventually live here; for now, every entry links to its original publication.
01 02 Production AI · Medium
AI 2025: The Frontline and The Circus
Agents, hype, and the reality of building production systems beyond the noise.
03 Model adaptation · Medium
A Definitive Guide to QLoRA: Fine-tuning Falcon-7b with PEFT
A hands-on treatment of 4-bit quantization, LoRA configuration, and model adaptation with PEFT.
04 Retrieval · Medium
Understanding Retrieval-Augmented Generation: A Simple Guide
An early explanation of how retrieval and generation work together.
05 Model adaptation · Medium
QLoRA: A Leap Forward in Efficient Fine-Tuning of LLMs
Why NF4, double quantization, and paged optimizers changed efficient model adaptation.