VBT Task Author-AI Data Annotator

Hello,

Greetings from ZettaMine!!!

Freelancing Opportunity: We are hiring VBT Task Authors for a short-term AI Research, Benchmark Task Creation, and Red-Teaming project focused on evaluating and identifying weaknesses in advanced AI models.

Eligible Locations: Bangladesh, India, Indonesia, Egypt, Ghana, Nigeria, Turkey, Brazil, and Colombia.

Role: VBT Task Author-AI Research Analyst

Experience: 3+ Years preferred

Mode: Remote

Commitment: Full-time (40 hours/week) or Part-time (20 hours/week)

Working Hours: Minimum 4 hours of overlap with PST working hours.

Mandatory Eligibility Criteria

Technical Qualification:

• Strong analytical, problem-solving, and investigative research skills.
• Intermediate-to-advanced web research and Open Source Intelligence (OSINT) capabilities.
• Ability to discover hard-to-find facts using academic archives, public records, local databases, and other reliable sources.
• Experience in fact-checking, AI benchmark task authoring, red-teaming, investigative research, or technical quality assurance.
• Familiarity with AI benchmarks such as GAIA, BrowseComp, SimpleQA, SWE-bench, Terminal-Bench, or Humanity’s Last Exam (HLE).
• Ability to write clear, precise, and unambiguous research prompts with evidence-backed answer keys.
• Strong attention to detail when verifying sources, identifying contradictions, and documenting source URLs.
• Ability to analyze AI model outputs, search logs, and research trajectories to identify errors and reasoning weaknesses.

Educational Qualification:

• Mandatory: Bachelor's degree or higher in a relevant engineering domain.
• Strong research, technical analysis, and documentation skills are preferred.

Experience:

• Experience in investigative research, fact-checking, AI benchmark task creation, red-teaming, or technical QA review is preferred.
• Prior experience designing challenging research tasks that expose weaknesses in advanced AI models is an advantage.
• Familiarity with AI model evaluation frameworks and benchmark testing is beneficial.

Availability:

• Full-time commitment of 40 hours per week or part-time commitment of 20 hours per week.
• Minimum 4 hours of overlap with PST working hours.
• Ability to complete assigned research and benchmark tasks within the project timelines.
• Availability for the full 4-week engagement.

Others:

• Ability to work independently in a remote environment.
• Strong written communication and documentation skills.
• Ability to validate information using primary or highly credible sources.
• Experience with prompt-injection or reasoning-trap tasks is an advantage.
• Prior experience as a benchmark reviewer, auditor, or quality lead is preferred.
• Reliable computer/laptop and stable internet connection.

What You’ll Work On

• Research niche topics to discover difficult-to-find facts that challenge advanced AI models.
• Verify factual information using primary sources and credible references.
• Develop clear, evidence-based benchmark tasks designed to test AI research and reasoning capabilities.
• Create precise answer keys supported by reliable evidence and exact source URLs.
• Test benchmark tasks against designated state-of-the-art AI models to identify potential weaknesses.
• Analyze model outputs, search logs, and intermediate results to determine where and why failures occur.
• Distinguish genuine model limitations from ambiguous prompts or insufficient evidence.
• Identify recurring failure patterns and apply them across different research domains.
• Review tasks created by other contributors to ensure accuracy, clarity, and quality.
• Maintain consistent documentation and follow project evaluation guidelines.

Perks of Freelancing

• Fully remote work environment.
• Flexible full-time or part-time commitment.
• Opportunity to contribute to advanced AI training and evaluation projects.
• Hands-on experience with AI benchmarking, investigative research, and red-teaming.
• Opportunity to help improve the research and reasoning capabilities of next-generation AI models.

Interested candidates are requested to share their updated CV with us at md.afreen@zettamine.com

Thanks & Regards,

Saranya M (Afreen)

ZettaMine