Apple's Aug 21 Cut: The ISE Sourcing Gap Nobody Is Searching

Every recruiter in the Valley is now typing "ex-Apple Siri" into LinkedIn after Bloomberg's August 21, 2026 report that Apple cut more than 200 people across Vision Pro, Siri, and a third software org almost nobody is Boolean-searching for. That is the wrong query, and it is why the best on-device AI engineers in this cohort will still be reachable next week.

The team you actually want is called Intelligent Systems Experience, or ISE. It is small, it is misspelled in Apple's own job postings, and it is where the Foundation Models plumbing lives.

What Apple actually cut on August 21

Apple laid off roughly 200 people across three teams: about 100 from Vision Pro (gaming and immersive video), and about 100 from Siri plus a software group called Intelligent Systems Experience. The Siri portion targeted staff working on the pre-LLM assistant, not the new Siri AI system Apple demoed on June 8, 2026 for developer testing.

The three-team breakdown, from Bloomberg and Engadget reporting:

No WARN filing covers these cuts. Laid-off Apple staff will not appear in any public WARN database, so recruiters relying on state filings to build a list will get nothing. You have to source directly, and you have to know what to search for.

Why "Intelligent Systems Experience" is the sourcing key

ISE is the on-device AI integration group inside Craig Federighi's software engineering org, and it is where Apple builds the plumbing that runs Foundation Models on the device. Sebastien Marineau-Mes was named VP of ISE under Federighi in 2020 and has led it since. Almost no external recruiter searches for the team by name.

Three sub-teams sit inside ISE, each with a different profile:

• Apple Intelligence Platform. Builds the on-device software infrastructure powering Writing Tools, Siri, Visual Intelligence, and Image Playground. This is the Foundation Models API team.
• ISE Engineering. Owns integration and tooling across iOS, iPadOS, macOS, and visionOS, plus core system apps: Setup, Settings, Files, Finder, Spotlight, Contacts.
• ISE Incubation. A Cupertino prototyping unit that combines platform engineers, ML engineers, and designers to turn ideas into shippable demos.

If you are building a local LLM on constrained hardware, the Apple Intelligence Platform sub-team is the highest-value pool in the entire ex-Apple diaspora. There are not many people on Earth who have shipped a production on-device Foundation Models API. Most of them were, until last week, sitting inside ISE.

The Boolean typo that hides half the team

Apple's own job postings spell the team name two different ways, and this is the single most exploitable arbitrage in the story. Bloomberg and Apple's official comms use "Intelligent Systems Experience" (plural). Multiple live Apple job listings on themuse.com use "Intelligent System Experience" (singular).

If you Boolean-search only the plural form, you miss roughly half of the ISE-alumni profiles that describe themselves using the singular. That means the two most likely queries a competing sourcer runs both fail:

1. "ex-Apple" AND "Siri" returns the wrong cohort entirely (grading analysts, not platform engineers).
2. "Intelligent Systems Experience" returns about half the real ISE population.

The correct query is an OR across both spellings, plus the sub-team names, plus adjacent surface areas like Spotlight, Foundation Models, and Writing Tools. That is not a query most people are going to hand-build. It is the exact gap Refolk closes: describe the person in plain English ("on-device LLM infra engineer from Apple's Intelligent Systems Experience team, either spelling") and get a ranked shortlist across GitHub, LinkedIn, and the open web.

263

The numbers behind the pool

In Refolk's index, the US "Apple Siri" pool is 263 people and only 22 of them currently list Apple as their employer. That is an 8.4% in-seat rate, which tells you the ex-Apple Siri diaspora has been leaking for a while.

The top titles in the "Apple Siri" pool: Siri Engineer, Siri Grading Analyst, Senior Natural Language Software Engineer, Machine Learning Engineer. Geographic concentration: Cupertino, Sunnyvale, Los Gatos, San Jose, and the broader Bay Area. If your role is remote-first or coastal-flexible, understand that the effective pool for you is a fraction of 241, because Cupertino gravity is real and Apple lifers do not relocate easily.

The Siri cohort is a skills mismatch, not a talent windfall

The Siri cuts targeted teams working on the old version of the assistant, per Engadget, because the new Siri AI system requires different expertise. These are not laid-off LLM engineers. They are classical NLU engineers, rule-based dialog specialists, and grading analysts.

Who should hire them, and who should not:

• Good fit: automotive voice interfaces, industrial voice, accessibility tech, any product where deterministic dialog matters more than open-ended generation.
• Good fit: eval and annotation teams at frontier labs, where Siri grading analysts have exactly the muscle memory needed.
• Poor fit: LLM-native startups looking for people who have shipped transformer inference at scale. You will bounce off this pool and burn six weeks doing it.

This is where sourcers who pattern-match on "ex-Apple AI" without reading past the headline lose the most time. If you are an LLM-native founder, skip the Siri cohort and go directly for ISE Apple Intelligence Platform alumni.

What ISE alumni actually know how to build

ISE engineers built on-device inference infrastructure, Foundation Models APIs, and the orchestrator that routes user intent between local models and Private Cloud Compute. That skill set is the rarest in the entire ex-Apple diaspora and the most valuable for a specific set of buyers.

The Siri AI architecture Apple announced on June 8, 2026 tells you exactly what these engineers touched:

1. On-device Apple Foundation Models. Quantization, memory-mapped weights, Neural Engine scheduling.
2. Private Cloud Compute. Attested inference, secure enclave to server handoff, request routing.
3. System orchestrator. Deciding what runs locally vs. in cloud, latency budgeting.
4. Spotlight index integration. Semantic search over personal context.
5. App Toolbox. Cross-app action invocation, the thing that makes an assistant actually do something.

A hardware startup shipping a local LLM on custom silicon needs items 1, 3, and 5. A frontier lab shipping consumer assistants needs items 2, 4, and 5. An enterprise agent company needs item 5 more than anything else. These are not interchangeable engineers, and a generic "ex-Apple AI" pitch will fail. You need to target by sub-team.

The Aug 21 cut is not a Siri story. It is a plumbing story, and the plumbers are outnumbered by the recruiters chasing them.

The exec-move leading indicator you already missed

Apple's Aug 21 staffing cut was the visible consequence of a reorg that had already happened on paper by April 2026. John Giannandrea, previously SVP of ML and AI Strategy, left Apple in April 2026 after Siri had already moved out of his remit. Amar Subramanya took the AI VP role, arriving from Microsoft (where he was corporate VP of AI) after 16 years at Google leading engineering for Gemini Assistant.

Anyone tracking exec moves as leading indicators had a four-month head start on this pool. The pattern is worth internalizing:

• A senior AI exec departs.
• Their org gets restructured under a new leader with different priorities.
• Three to six months later, staff whose skills do not match the new mandate get cut.

Subramanya's Gemini Assistant background telegraphs an LLM-native, cloud-hybrid direction. Classical NLU engineers on legacy Siri were never going to fit that. The cut was legible in April. Sourcing tools that only react to news are three months late; the ones that let you standing-query a rolling watchlist of Apple ML engineers let you build relationships before the layoff even lands.

How to source ISE alumni this week

Do this in the next seven days, before the second wave of recruiters figures out the team name:

1. Query both spellings. Build a saved search that ORs "Intelligent Systems Experience" and "Intelligent System Experience." If your tool cannot handle typo-tolerant OR queries across LinkedIn and GitHub, use one that can.
2. Filter by sub-team. Apple Intelligence Platform, ISE Engineering, ISE Incubation. Prioritize by which sub-team maps to your product.
3. Exclude the Siri grading cohort unless you specifically want eval talent. Grading Analyst titles skew toward annotation work.
4. Check GitHub for Core ML and Foundation Models sample code contributions. Anyone who filed issues against Apple's ML repos in 2025 or 2026 is worth a look, whether or not they were laid off.
5. Map their commute radius. If your role is not in Cupertino, Sunnyvale, or Los Gatos, filter for candidates who have moved before or list a non-Bay location on side projects.
6. Reach out with a sub-team-specific pitch. "I saw you worked on the Apple Intelligence Platform Foundation Models API" beats "ex-Apple AI engineer" by a factor of ten on reply rate.

Apple's restraint is why this pool matters

Apple almost never lays off engineers, so when it cuts 200 the rarity is the point. For context: Oracle shed 21,000 roles in its fiscal year ending May 2026, Amazon announced roughly 16,000 corporate cuts, Microsoft cut about 4,800, and Cisco planned around 4,000. Apple cut 200.

The ISE cohort inside that 200 is small enough that a founder with a targeted list and a specific pitch can realistically hire two or three of the best people in the world at what they do. Do not squander that by searching for "Siri."

FAQ

How many ex-Apple AI engineers are actually reachable after the Aug 21 cut?

In Refolk's index, the broad "Apple Siri" US pool is 263 people, of whom only 22 currently list Apple as their employer. That puts roughly 241 ex-Apple Siri-adjacent engineers in-market before the Aug 21 cut, with another portion of the ~100 Siri and ISE headcount now joining them. The ISE-specific pool is materially smaller than the broad Siri number and disproportionately concentrated in the Bay Area.

Why does the "Intelligent Systems Experience" team name matter for sourcing?

ISE is Apple's on-device AI integration group, and its engineers built the Foundation Models APIs, orchestrator, and system infrastructure behind Siri AI, Writing Tools, and Visual Intelligence. They rarely list "Siri" or "Vision Pro" on their LinkedIn headlines, so standard ex-Apple AI queries miss them. Because Apple itself uses both "Intelligent Systems Experience" and "Intelligent System Experience" in job postings, any Boolean search that omits the singular variant loses about half the population.

Should I hire the laid-off Siri engineers for an LLM startup?

Probably not the classical NLU and grading cohort. Engadget's reporting confirms the Siri cuts targeted teams working on the pre-LLM assistant, because the new Siri AI system requires different expertise. Those engineers are excellent fits for automotive voice, accessibility tech, industrial voice, and frontier-lab eval teams, but a mismatch for LLM-native startups looking for transformer inference or RLHF experience.

What is the fastest way to build a targeted list of ISE alumni?

Use a sourcing tool that lets you describe the person in plain English rather than hand-building Boolean strings. Refolk searches across GitHub, LinkedIn, and the open web, handles both spellings of the team name in one query, and returns a ranked shortlist filtered to the sub-team that matches your product.