1. The "Agentic" Evolution vs. Simple RAG

We are seeing a massive pivot from basic RAG (Retrieval-Augmented Generation) setups to Agentic Workflows.

Startups are no longer looking for "Prompt Engineers"; they are hunting for architects who can build autonomous loops, handle tool-calling reliability, and manage multi-agent orchestration.

The Data: 60% of our current AI-related mandates specifically require experience with frameworks like LangGraph, CrewAI, or AutoGen.

​

2. Efficiency over Brute Force (The Token Economy)

The "Burn Rate Awareness" has hit the R&D floor. Founders are prioritizing engineers who can optimize for Inference Costs and Latency. Knowledge of quantization, small language models (SLMs) for specific tasks, and prompt caching is becoming a non-negotiable requirement for senior roles.

The Vibe: Being "AI-capable" is no longer enough; being "AI-Efficient" is the new gold standard.

​

3. The Local LLM Movement

Data privacy and IP security are driving more Israeli startups (especially in Cyber and FinTech) to move away from pure OpenAI dependency. We see a 35% increase in demand for talent capable of deploying and fine-tuning local models (Llama 3.x, Mistral) on private VPCs.

​

4. The "Vetting Gap" Reality Check

Our internal screening shows a stark reality: While 8 out of 10 candidates list "AI Expertise" on their CV, only 2 out of 10 can successfully pass an implementation-level live coding session involving complex vector database indexing or handling model hallucinations in production.