Innodata's 550 Have No Payslips. Verify Them Through EPFO and COEs

Hundreds of trained AI data-ops workers from Innodata in India and the Philippines are suddenly available, and almost none of them can give you the documents a normal background check asks for. Their payslips are stuck behind a revoked HR portal, their relieving letters wait on asset return, and most haven't posted anything on LinkedIn. If you hire annotation leads and QA staff, this is the best sourcing window of the quarter. To use it, you need a way to verify people that doesn't depend on Innodata's paperwork.

What happened at Innodata on Oct 26?

Innodata, a US-based data-engineering and AI-services company, laid off nearly 550 employees in India, with Oct 26 recorded as their final working day. In the Philippines it dismissed raters, quality experts, team leaders and operational managers on a Zoom call. The stated reason was "a reduction in business assignments", not individual performance.

The details matter for anyone verifying these candidates:

• Payslips had already been generated with Oct 26 as the final working day. Workers couldn't download them because their HR portal access had been revoked.
• Termination letters said final settlements would be processed once company assets were returned.
• Affected India staff set up a WhatsApp group that had nearly 500 members by late evening.
• The stock fell 26.4% as of 12:26 p.m. ET the next day.

The likely mechanism is client concentration. Innodata's 10-Q for the quarter ended June 30, 2026 shows one customer at about 37% of revenue and another at about 34%. Innodata hasn't said which customer or project was involved. The same filing says Q2 direct costs rose mainly because of "an increase in headcount to support higher volumes of AI data engineering services." Read together, it looks like many of these people were hired for volume that then went away. The cuts are more than 5% of the company's workforce.

For a hiring manager, the takeaway is simple. This was lost work, not a performance cull. Don't screen these candidates as if it were.

It does break the usual Indian background check, which leans on relieving letters and recent payslips. Background verification without a relieving letter means confirming employment history from records the former employer doesn't control. In India that's the Employees' Provident Fund Organisation (EPFO). In the Philippines it's the Certificate of Employment (COE). Both are covered below. First, you have to find the people.

Why does searching "Innodata" on LinkedIn find the wrong people?

In India, most profiles that mention "Innodata" don't belong to people who work there. In Refolk's index of professional profiles, a headline keyword search for "Innodata" returns 7,609 India profiles. Only 1 of 25 sampled profiles listed Innodata as the current employer, and most of the rest were founders or CEOs.

The Philippines looks very different. The same search returns 1,482 profiles there, and 5 of 25 sampled listed an Innodata entity (Innodata Inc. or Innodata Knowledge Services, Inc.) as their employer. Those profiles cluster in Metro Manila and Cebu.

A keyword count tells you how often a word appears. It doesn't tell you who worked somewhere. In the Philippines, the local legal entity shows up as the employer often enough that the name search pays off. In India, real employees are buried under unrelated matches, so a name search mostly wastes time. Search by skill and title instead.

How big is the hidden data-ops pool?

It's much bigger than online profiles suggest. Refolk's index has 207 India profiles that combine Data Annotation as a skill with a Team Lead, Quality Analyst, Delivery Manager or Project Manager title. Innodata's India cut alone is about 2.7 times that number.

2.7x

One caveat: the 550 includes people who aren't leads. Even so, the gap is large. Here's what the visible 207 look like:

• Titles (sample of 25): Quality Analyst (10), Project Manager (5), Team Lead (5).
• Employers in the sample: Wipro (4), Accenture (2), Cognizant (2), TELUS Digital (2), iMerit (1).
• Geography: Hyderabad is the largest cluster.

Why is this layer so hard to see? The people in it don't brand themselves online. When 550 of them lost their jobs on the same day, nobody posted on LinkedIn. They joined a WhatsApp group.

For a recruiter, this means a cold search for an annotation team lead drags on. And when it ends, you're looking at the same couple of hundred visible profiles that every other AI-data employer is already calling. The experienced leads who just became available are mostly not on that list.

Where do you actually find Innodata's leavers?

Look through people, skills and neighboring employers, not the company name. The fastest channels, in order:

1. The WhatsApp group, through a referral. Find one former Innodata employee you trust and ask them to share your role. Don't scrape the group or cold-add people. A role posted by a peer gets read. A stranger's message gets reported.
2. Skill-plus-title searches in India. Data Annotation plus Quality Analyst, Team Lead or Project Manager, starting in Hyderabad.
3. Neighboring employers. iMerit, TELUS Digital, Wipro, Accenture and Cognizant are the top employers for annotation leads in Refolk's India index. That's the adjacent talent pool Innodata's leavers sit in.
4. Entity-name searches in the Philippines. Search for "Innodata Knowledge Services, Inc." in Metro Manila and Cebu. With a 20% hit rate, the name search is worth doing there.

Channels 2 and 3 are the tedious part, because they mean running several title variations across a handful of employers. Refolk takes that as one plain-English request:

Once you have names, the WhatsApp group is the best place to check them. Ask your referrer whether each person really led a team. Then do the formal verification.

What does EPFO service history verification prove?

EPFO service history verification confirms who employed a candidate and exactly when. It shows the legal employer names, the EPFO establishment IDs, and the date of joining and date of exit for every period of service. It requires the candidate's consent. It doesn't confirm performance, responsibilities, salary or the reason for leaving.

The record is tied to the worker's Universal Account Number (UAN) and held by the government, not by Innodata, so a locked HR portal doesn't affect it.

You can ask the candidate to share their service history directly, or use a UAN verification vendor such as Cashfree's Employment 360 or Digiverifier. Both need consent. For a candidate who has just lost their job, a clear consent request that names the vendor and says what you'll see builds trust.

Why is an "ongoing" Innodata entry not a red flag?

The worker can't close it yet. The EPFO portal blocks a worker from marking their own exit until two months after the employer's last PF deposit. Until roughly late December, an Innodata leaver will still show as employed there.

A screener who doesn't know this will flag the candidate as "still employed, possible dual employment" and move on, which throws away the best candidates in the batch. Check these instead:

• Establishment ID: does it match Innodata's?
• Contribution history: is it continuous through the candidate's stated tenure?
• Last contribution month: does it fit an October exit?
• Joining date: does it match what the candidate told you?

A missing exit date tells you nothing during this window. A gap in contributions in the middle of a claimed tenure is worth a question.

Why is the Philippines easier to verify?

Workers in the Philippines are entitled to a Certificate of Employment within three days of asking for it. Pending accountabilities, such as unreturned laptops, shouldn't automatically block a factual COE. The COE also states the type of work performed, which EPFO never does.

3 days

So for the raters, quality experts and team leads cut in Metro Manila and Cebu, ask the candidate to request a COE from Innodata Knowledge Services, Inc. right away. It costs them nothing, it puts a date on the clock, and it gives you something EPFO can't: the employer's own description of the work. In India, the equivalent paper (the relieving letter) waits on asset return, so verification there is mostly up to you.

How do you verify a "team lead" claim?

Government records can't do it. EPFO proves how long someone was employed, not how senior they were, and without payslips or appraisal records the title "team lead" has nothing behind it. Pair the EPFO check with a short work test and a peer reference.

Inter-annotator agreement (IAA) is how consistently different annotators label the same items. Calibration is the process a lead runs to raise it. Both are what a real data-ops lead does every day, so test them directly:

• Calibration task: give them a small set of labeled items where annotators disagree. Ask them to settle the disagreements and write the guideline change that would prevent them.
• IAA reasoning: ask how they'd measure agreement on that set and what score would make them retrain a rater.
• Escalation scenario: a client changes the labeling rules halfway through a batch. What do they do with the work already delivered?
• Peer reference: two people from the WhatsApp group who reported to them, not one manager.

EPFO tells you they were there. A calibration task tells you whether they were the one everyone checked their work against.

Someone who has actually run a QA floor will settle the disagreements and argue with your guideline. Someone who has only done the labeling will relabel the items and stop there.

A five-step playbook for this week

The candidates are available now, and the formal paperwork won't show up for weeks. Run this sequence:

1. Source by skill, not company name. In India, search for Data Annotation plus lead or QA titles, starting in Hyderabad. In the Philippines, search the Innodata entity name in Metro Manila and Cebu.
2. Get one referral into the WhatsApp group. Post the role through a peer, with the salary range and the interview steps.
3. Collect consent and run EPFO or COE. Use the candidate's own record or a vendor like Cashfree or Digiverifier in India. Ask Philippine candidates to request their COE on day one.
4. Read EPFO correctly. Check the establishment ID and continuous contributions. Ignore the missing exit date until late December.
5. Verify seniority with work, not paper. Calibration task plus two peer references, then make an offer.

Firms that wait for relieving letters will be weeks behind. Firms that verify through EPFO and COEs can have a team in place first.

If you're benchmarking your offer against what other employers are hiring for right now, the operations market in Bangalore is a useful reference for scope and titles.

FAQ

Can I run EPFO service history verification without the candidate's consent?

No. UAN-based verification needs the candidate's consent, whether you ask them to share their own service history or go through a vendor such as Cashfree's Employment 360 or Digiverifier. Ask plainly, name the vendor, and explain what you'll see: employer names, establishment IDs and dates of joining and exit. Nothing about salary, performance or why they left.

Are the Innodata layoffs a performance signal?

No. Termination letters gave "a reduction in business assignments" as the reason, not individual performance. With two customers making up about 37% and 34% of revenue in the June 2026 quarter, losing work from a single client can plausibly explain cuts at this scale. Treat these candidates like any team affected by a project ending.

When will Innodata India leavers get relieving letters?

It's unclear. Termination letters tied final settlements to the return of company assets, and nothing public gives a date. Don't make relieving letters a condition of your offer. Make the offer conditional on EPFO verification and a satisfactory work test instead.