TikTok's Bellevue WARN: The US E-Comm ML Pool Is 26 People

If you recruit for e-commerce ranking, ads, or seller-growth ML, you have spent this year trying to build a shortlist that does not exist in any keyword search. TikTok's late-August 2026 WARN filing in Washington just named 75 of the people you were looking for.

Most coverage is treating the notice as another routine tech cut. It is not. It is a named window into one of the smallest sourceable pools in US tech, and the timing rules for approaching it are the opposite of what most sourcers will run.

What the WARN actually says

TikTok is cutting 75 workers at Lincoln Square North, 700 Bellevue Way NE, with a hard separation date of October 19, 2026. The filing entity is TT Commerce & Global Services LLC on TikTok letterhead, listing two ByteDance employees as contacts and copying Bellevue Mayor Mo Malkoutian.

The roles named in the notice cluster in five buckets:

• Anti-fraud and governance program managers
• Seller and creator operations staff
• Campaign managers
• Data scientists
• Backend and frontend engineers

TikTok's stated rationale in the notice: the cuts are "necessary given recent restructures to the Company's operations." Affected employees got at least 60 days' notice, can apply for internal transfers through the October 19 date, and will receive two months' severance if they don't land one.

Two other facts matter for scale. TikTok employs roughly 1,600 people in the Puget Sound region, so this is a surgical 4.7% cut, not a site closure. And this is the second consecutive year of TikTok layoffs in Bellevue: July 2025 saw a combined 65 people cut across separate TikTok and ByteDance WARN filings, most of them also TikTok Shop. That gives you a nameable 140-person 2025 plus 2026 Bellevue TikTok Shop alumni cohort to work with.

Why 75 layoffs redraws a national pool

The visible US pool of e-commerce data scientists, MLEs, and applied scientists is only 26 people, and TikTok is already the single largest employer inside it. A 75-person Bellevue cut, even if only a fraction are DS or MLE lines, redistributes something like 15 to 20% of the nationally visible e-comm ML cohort in a single filing.

That number is not a typo. In Refolk's index of professional profiles, when you filter for US-based data scientists, MLEs, and applied scientists whose work is explicitly tagged to e-commerce, you get 26 names. Not 260. Not 2,600. Twenty-six. TikTok employs 5 of them.

26

Here is the full slice from Refolk's index for the query most sourcers would run on this WARN.

The point isn't that only 26 e-commerce ML people exist in the US. The point is that only 26 self-describe that way clearly enough to be findable by keyword. Everyone else is hiding inside "recommendation systems at $BIG_CO" or "search relevance" or "ads ranking." The WARN's value to a recruiter is that it names 75 people whose LinkedIn headlines are about to update with "ex-TikTok Shop," which is the exact keyword that unlocks the hidden portion of the pool.

The realistic buyers are 4 companies, not 40

Four to five companies actually compete for these candidates: Moloco, Instacart, eBay, Amazon, and The Home Depot's ML org. Every one of them appears in the same e-commerce ML cohort in Refolk's index. Samsung Electronics America shows up as a wildcard sixth.

Split them by what they'll actually buy:

• Amazon buys the SWEs, both backend and frontend. High volume, mid-signal outreach wins.
• Moloco is the sleeper buyer for the ranking and seller-growth scientists. Its entire business is ML-driven commerce ranking and ads. The mapping to TikTok Shop work is 1:1.
• Instacart buys ads ranking and recommendation scientists. Similar surface area, different vertical.
• The Home Depot has a quietly serious ML org for search and merchandising. Underrated destination for staff-level ICs.
• eBay buys generalist e-commerce DS, especially anyone who's worked on trust, fraud, or seller quality.

If you're a founder reading this and your product is not one of those categories, this WARN is not your pool. The instinct to treat every big-name layoff as "our chance" burns weeks. The instinct to match the affected sub-team to the two or three companies whose product actually rhymes with it will get you a hire.

The WARN doesn't create a pool. It renames a hidden one so keyword search can finally see it.

What about the Nashville 250?

Do not conflate them. Earlier in August, TikTok announced the closure of its Nashville office, eliminating about 250 positions. Those roles were content moderation, not ML. Different candidate profile, different buyers (Discord, Roblox, and Character.AI for Trust and Safety), different pitch. If you're an e-commerce ML recruiter, the Nashville notice is noise. If you're a T&S recruiter, it's the month's biggest event and it's already been undercovered.

The severance window inverts normal timing

The right time to send outreach is the first two weeks of October and again mid-November, not September. Here is the sequence.

Affected employees can apply for internal transfers all the way through the October 19 separation date, and they receive two months of severance if they don't take one. That means:

1. Early-to-mid September: candidates are actively interviewing internally. Cold outreach lands in a spam folder in their head. Skip.
2. Late September: internal-transfer decisions start coming back. Some yes, mostly no.
3. First two weeks of October: transfer denials are fresh. Severance is starting. This is the highest-response window of the entire cycle.
4. Mid-October to mid-November: candidates are decompressing. Response rates dip.
5. Mid-November onward: severance runway visible. Second peak.

The recruiters who blast in August will burn the list before the candidates are psychologically available. The recruiters who wait, build the named list now, and sequence outreach for October 6 to October 17 will win the top three names.

Building that list is the friction that eats a week of manual work. You need to filter by "TikTok Shop" plus Bellevue plus DS or MLE titles, cross-reference against the July 2025 alumni cohort so you don't waste messages on people who already left, and rank by seniority. That is the exact gap Refolk closes: you describe the pool in plain English and get a ranked shortlist across LinkedIn, GitHub, and the open web.

The 2025 alumni cohort is your proof point

The single most useful sourcing artifact from this WARN isn't the 75 names. It's the ability to cite the 65 people who were cut in July 2025 and show a candidate where they landed. That turns "TikTok Shop instability" from a rumor into a datapoint.

Two consecutive years of Bellevue cuts change the candidate narrative in a way one-off layoffs never do. The pitch to a wavering data scientist stops being "come to my startup" and becomes "here are your former colleagues, here's where each of them is now, and here's why the ranking and ads people clustered at Moloco and Instacart." That is a conversation. A cold InMail about "exciting opportunity" is not.

Candidates believe named cohorts, not vibes. Recruiters who can produce the 2025 cohort's destination map on request outperform recruiters who cannot on this specific WARN. In Refolk you'd query for people who worked at TikTok in Bellevue between January 2024 and December 2025 and no longer work there, pull the list, group by current employer, and screenshot it. That artifact is worth more than a week of cold outreach.

Regional context: no one nearby is hiring the overflow

Local absorption capacity is close to zero. Zillow cut more than 500 jobs this month including 91 in Washington state. Microsoft eliminated 605 positions in the state in July as part of a broader 4,800-person reduction. Google cut 52 jobs locally this month and Salesforce cut 59. The macro floor doesn't help either: tech layoffs in 2026 surpassed all of 2025 by August 6, with 125,759 employees affected across 264 companies according to Layoffs.fyi.

125,759

Practical consequence: the affected TikTok Shop candidates will look outside Seattle. The Bay Area competitors (four visible in the cohort) become the realistic destination, along with remote roles at Moloco and Instacart. If you're a Seattle-based startup hoping the WARN gifts you five local hires, expect two, and expect to compete against remote offers from the Bay for the good ones.

What to do this week

The named list matters more than the pitch. Do these five things in order:

1. Pull the WARN attachment. Confirm the exact DS and MLE headcount inside the 75. Public reporting groups titles; the filing itself breaks them out.
2. Build the 140-person alumni cohort. 2025 Bellevue TikTok cuts (65) plus the 2026 WARN (75). Deduped, current employer attached, seniority tagged.
3. Map to the four buyers. For each candidate, tag which of Moloco, Instacart, Amazon, eBay, or Home Depot they'd realistically move to. If none, they are not your candidate.
4. Draft outreach for October 6. Not August, not September. Write it now, schedule it then.
5. Prep the 2025 destination map as an attachable one-pager. This is your differentiator when the candidate has eight InMails from other recruiters that day.

FAQ

How many of the 75 TikTok Bellevue layoffs are data scientists?

Public reporting groups the affected roles into buckets (anti-fraud PMs, seller and creator ops, campaign managers, data scientists, backend and frontend engineers) without publishing a per-title count. The WARN attachment filed with Washington's Employment Security Department contains the itemized breakdown; a public records request or the state's WARN portal will give you the exact figure. Expect DS and MLE to be a minority of the 75.

When is the best time to reach out to affected TikTok Shop employees?

The first two weeks of October 2026, then again in mid-November. Employees can apply for internal transfers through the October 19 separation date and receive two months' severance, so September is dominated by internal interviews and low response rates. Transfer denials come back in late September, which makes early October the highest-signal outreach window. Mid-November catches the candidates whose severance runway is getting visible.

Which companies will actually hire TikTok Shop's ML people?

Realistically four to five: Moloco, Instacart, Amazon, eBay, and The Home Depot's ML org, with Samsung Electronics America as a wildcard. All appear in the same US e-commerce ML cohort in Refolk's index. Amazon takes the SWEs at volume. Moloco is the sleeper for ranking and ads scientists because its business model is 1:1 with TikTok Shop's ML surface area. Instacart competes for ads and recommendation ICs. eBay picks up trust, fraud, and seller-quality DS. Home Depot quietly hires staff-level search and merchandising ML.

Is the TikTok Nashville closure part of the same sourcing story?

No. The Nashville announcement (about 250 positions) covers content moderation, not ML or engineering. Different candidate profile, different destination employers (Discord, Roblox, Character.AI for Trust and Safety work), different pitch. If you recruit for e-commerce ML, treat Nashville as noise and focus on the Bellevue WARN. If you recruit for Trust and Safety, Nashville is the month's most important event and it has been under-covered relative to its size.