Insights
Notes from production.
Engineering notes on voice agents, AI agents and LLMs in production, by Dr. Yochai Edlitz and the MLAIA team that builds Ultrabot.
All posts
RAG · August 2026Enterprise RAG That Ships: The Five Failure Modes That Kill PilotsWhy RAG pilots stall between demo and production, and a 90-day plan to ship.
On-prem AI · August 2026Sovereign AI in Israel: Running LLMs On-Prem in Regulated IndustriesData residency rules, the capability trade-off and a reference architecture for on-prem LLMs.
LLMs · Reference · July 2026LLMs in ProductionPrompting vs RAG vs fine-tuning, layered evaluation, hallucinations, SLMs and cost.
Voice AI · July 2026Voice AI Grows Up: Engineering Production Speech Systems in 2026Cascaded vs speech-to-speech, the sub-500 ms latency budget, per-call costs and Hebrew.
AI agents · June 2026The Agentic AI Production Gap: Why Most Enterprise Pilots Never ShipFive root causes that kill agentic pilots, and what the teams that ship do differently.
On-prem AI · June 2026Why Israeli Enterprises Are Bringing AI On-Prem with Small Language ModelsThe economics, privacy gains and production playbook for small language models.
AI agents · May 2026From Pilot to Production: Why 88% of AI Pilots StallIntegration, non-determinism, scope and governance: five principles for production agents.
AI agents · April 2026Agentic AI in Production: A Practical Framework for Israeli EnterprisesScope design, tools and memory, human-in-the-loop patterns and observability.
Talk to a person
Tell us what
you’re working on.
Share the problem, the data you have and what success would look like. We’ll discuss a practical next step.
Please don’t include confidential datasets, credentials or patient information.