What we do

Zefr is the leading global technology company enabling responsible marketing in walled garden social environments. Our solutions let brands manage where their advertising shows up on YouTube, Meta, TikTok, and Snap, in line with industry-standard frameworks. Through patented AI technology, Zefr gives brands and agencies more accurate and transparent tools for social walled gardens. We're headquartered in Los Angeles, with additional locations across the globe.

What you'll do

Every AI system that can tell a cooking video from a combat video learned it from a person who made that call first. That person is you.

We're hiring a Multilingual Data Annotation Specialist to label video, image, and metadata content across YouTube, TikTok, Meta, and Snap, and to bring cultural fluency to calls that a model can't make on its own. A clip that reads as harmless in one market reads very differently in another. Catching that difference is the whole job.

Your labels become the ground truth our machine learning systems train on and get measured against. When you get a hard call right, the system gets better at it. When the guidelines don't cover the case in front of you, you're the one who tells us the guidelines need to change.

This role suits someone who finds ambiguity interesting rather than frustrating, who'd rather ask a sharp question than guess, and who takes real satisfaction in precision.

Here's what you'll get to do

Label social media content against Zefr's contextual category guidelines using our in-house tooling.

Make daily judgment calls on video, image, and metadata — including content where cultural and linguistic context changes the answer.

Apply labeling frameworks, then help us sharpen them when you find the gaps.

Spot emerging trends and edge cases early, and raise them before they become systemic mislabels.

Run root cause analysis on quality misses and turn what you find into process changes.

Perform lingual quality assurance across multiple project pipelines.

You'll also build a working understanding of machine learning and computer vision, and of exactly how annotation quality shows up in model performance.

What we're looking for

Language

Oral, written, and cultural fluency in Hindi and English is required.

Experience and skills

Strong cultural fluency and a feel for context across varied content types.

Regular user of TikTok, YouTube, Facebook, Instagram, and Snapchat. You should know these platforms as a native, not a visitor.

Exceptional attention to detail, sustained across repetitive work.

Clear written communication, especially when explaining ambiguity, inconsistency, or definitional drift.

Comfort applying style guides and holding consistency across formats.

Comfortable using AI tools such as Claude, Gemini, or ChatGPT.

Proficiency in Google Workspace.

Nice to have

Background in annotation, content moderation, labeling, quality assurance, or policy enforcement.

SQL familiarity.

Additional languages — Spanish, Portuguese, Mandarin, Turkish, Tagalog, or others — are a plus.

How you work

Comfortable reviewing sensitive content, including political topics, religious perspectives, violence, and adult material.

Reliable and coachable. You take feedback well and apply it quickly.

Analytical, with structured problem-solving instincts.

Steady in a fast-paced environment with shifting priorities.

Effective both on a team and on your own.

Sound judgment with confidential and proprietary information.

Able to work a hybrid schedule, 2–3 days per week in our Marina del Rey office.

One thing to know going in

This role involves regular exposure to sensitive content, including political and religious material, violence, and adult content. It's a real part of the work, and we'd rather you know that now than find out later. We provide guidelines, support, and a team that talks openly about the hard cases.

Benefits (for US based employees)

Flexible PTO

Medical, dental, and vision insurance with FSA options

Company-paid life insurance

Paid parental leave

401(k) with company match