GUI / Computer-Use Videos Data Annotator

About the Company

Skyfall is working to reshape the AI landscape by creating the first enterprise world model. Its mission is to address major LLM limitations such as safety risks, hallucinations, and high training costs, while helping organizations better understand how data, people, and processes interact.

The founding team includes the creators of Maluuba, who helped pioneer the deep learning revolution. Before Maluuba was acquired by Microsoft for $160M and became Microsoft’s AI research center in Canada, the team collaborated with leading AI figures including Yoshua Bengio and Richard Sutton.

About the Role

This position focuses on building the training data that teaches the world model how people use software. The model learns from real screen recordings, so the accuracy of every curated label directly affects what it learns and how well it performs.

You will work with the world modeling team to review large sets of computer-use videos, validate AI-generated labels, distinguish genuine task workflows from non-workflow content, and identify noise or unusual cases. The work demands careful judgment, strong consistency, and a commitment to accuracy over speed, since even small mistakes can affect model training and downstream performance.

Key Responsibilities

• Check AI-suggested taxonomy labels such as Domain, Subdomain, and Software on sampled videos and mark each one as correct or incorrect based on the labeling rules.
• Watch video clips and decide whether they show real computer-use workflows, meaning step-by-step task execution, or non-workflow material such as commentary, passive browsing, or general discussion.
• Tag visual noise in sampled frames, including zooms, overlays, cursor effects, callouts, transitions, and similar elements, using one or more labels per frame as instructed.
• Handle shared, traceable annotation buckets across data sets that may contain anywhere from 1,000 to 5,000+ videos.
• Identify unclear cases, edge cases, or tooling problems and raise them with the team instead of making assumptions, while adding clear notes for escalation.
• Keep quality and productivity aligned with team targets and adapt quickly as the annotation rules change.
• Highlight missing, unclear, or ambiguous parts of the labeling guidelines and taxonomy so the process can improve over time.

Core Requirements

Attention to detail, judgment, and integrity

• Exceptional precision and the ability to stay consistent during long, repetitive, high-volume work.
• Ability to follow detailed labeling instructions closely and apply them consistently, even as they evolve.
• Good judgment to tell apart real workflows from non-workflow content and recognize edge cases.
• Strong integrity: accuracy must come before speed, and uncertainty should be escalated rather than guessed.

Software and technical fluency

• Strong general computer literacy and comfort using desktop operating systems such as Windows, macOS, or Linux, along with web browsers and common productivity, creative, and developer tools, so that real workflow behavior is easy to recognize.
• Comfort using web-based annotation platforms, spreadsheets, and tracking tools.
• Clear written English for documenting flags, notes, and edge cases in a precise way.

Qualifications

• A bachelor’s degree in any discipline; Computer Science, IT, or Engineering is an added advantage.
• 0 to 3 years of experience in data annotation, manual QA or software testing, IT operations, or another detail-heavy, rule-based role. Strong freshers with the right aptitude may also be considered.
• Ability to work full-time from the office at HSR Layout, Bengaluru for the full 3-month contract period.

Preferred Background

• Experience in AI data operations, video or content auditing, e-commerce or retail operations, or medical coding, where diligence and repetition are important.
• Basic understanding of how AI and ML models learn from training data.
• Familiarity with structured data formats such as labels, JSON, and taxonomies.

Who Will Do Well in This Role

• Someone who is methodical, patient, process-oriented, and dependable in repetitive, high-focus work.
• Someone who remains just as accurate and careful on the 400th video as on the 4th.
• Someone who treats ambiguity as a reason to pause and escalate rather than to make a guess.

What This Role Is Not

• This is not a software engineering, development, or programming position.
• This is not a creative, content-production, or design role.
• It is not suitable for casual computer users; it requires solid desktop-software fluency and comfort with repetitive, precision-based work.

Additional Information

This is a 3-month contract role based on-site in Bengaluru, at HSR Layout. The work requires full-time availability and a strong focus on quality, consistency, and careful judgment.