Online STEM Jobs in India (Remote, Full Time)

Rex.zone connects remote STEM professionals to AI/ML programs supporting large language model training pipelines. You will contribute to structured evaluation, training data quality improvements, and QA evaluation workflows across NLP, computer vision, and content safety labeling projects.

What You Will Do

* Execute rubric-based evaluations of LLM responses for accuracy, safety, and helpfulness

* Perform RLHF-style ranking, preference judgments, and prompt evaluation

* Validate data labeling outputs and enforce annotation guidelines compliance

* Conduct QA evaluation, audit sampling, and error analysis to find systematic issues

* Support NLP tasks such as named entity recognition (NER) and taxonomy alignment

* Support computer vision annotation tasks (e.g., bounding boxes, segmentation) when required

* Document edge cases and decisions to improve evaluation consistency

* Collaborate asynchronously with distributed stakeholders and meet delivery SLAs

Required Qualifications

* Mid-senior experience in engineering, data, analytics, or applied ML workflows

* Strong reasoning and written communication for prompt evaluation and scoring

* Practical Python and SQL for analysis, sampling, and QA checks

* Experience with QA evaluation, test case design, or data quality review

* Familiarity with ML concepts and dataset iteration/model performance improvement

Preferred Qualifications

* Experience with RLHF evaluation, human preference data, or LLM evaluation harnesses

* Exposure to data labeling platforms and reviewer workflows

* NLP experience (NER/classification/summarization evaluation) and/or CV annotation

* Knowledge of content safety labeling and policy-based evaluation

* Experience producing clear annotation guidelines and resolving edge cases

Compensation

Hourly range: $30–$50/hr (USD), based on scope and skills alignment.

How To Apply

Apply via Rex.zone with a resume highlighting STEM projects, Python/SQL, QA evaluation or review workflows, and any exposure to RLHF, prompt evaluation, data labeling, NLP, computer vision, or content safety labeling.