Meta cuts hundreds of jobs while launching KernelEvolve, its AI ranking engineer agent.

Meta’s new AI ranking‑engineer agent, KernelEvolve, is being rolled out as the company trims staff in the Bay Area, filings show. The system automates kernel authoring for Meta’s heterogeneous hardware fleet—NVIDIA and AMD GPUs, custom MTIA chips and CPUs—according to the internal engineering blog posted on April 2, 2026 (Engineering). By converting high‑level model operations into chip‑specific instructions, KernelEvolve aims to replace manual tuning by kernel experts, a process that “doesn’t scale” given the growing number of models and hardware generations (Engineering).

The layoffs affect 124 employees in Burlingame and 74 in Sunnyvale, with cuts effective in late May, state filings reveal (Nypost). Meta’s spokesperson said the company continues hiring for “critical roles” despite the reductions (Nypost). Headcount stood at 78,865 at the end of 2025, a 6 % year‑over‑year increase, even as the firm considers broader cuts that could exceed 20 % of its workforce (Nypost).

KernelEvolve is the second installment in Meta’s Ranking Engineer Agent series, which previously introduced autonomous model‑experiment design (Engineering). The new agent focuses on low‑level infrastructure, generating optimized kernels for custom operators beyond standard GEMMs and convolutions (Engineering). Its scope extends beyond Ads Ranking to any AI model that runs on Meta’s data centers, according to the blog authors.

The push reflects Meta’s strategic shift toward “machine‑driven systems,” analysts note, as the company pours $10 billion into a new data center in El Paso, Texas (Nypost). Recent weeks also saw roughly 700 Meta staff in operations, recruiting, sales and Reality Labs laid off, underscoring the breadth of the restructuring (Nypost).

Zuckerberg is reportedly weighing deeper cuts that could affect more than one‑fifth of Meta’s global workforce while the firm accelerates AI spending (Nypost). KernelEvolve, by automating kernel optimization, is intended to sustain performance gains as Meta scales its AI workloads, a critical component of the company’s broader AI investment plan (Engineering).