Saviour D.

Key Skills

Software

Other

Top Subject Matter

AI training data annotation (text classification, intent, NER-related labeling, sentiment/toxicity)

RLHF / LLM evaluation (preference data, scoring, and ranking)

Prompt authoring and adversarial testing for LLM training/evaluation

Top Data Types

Text

Top Task Types

Classification

RLHF

Prompt + Response Writing (SFT)

Freelancer Overview

Self-Directed Text Classification and Tagging Training and Practice. Brings 3+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Other. Education includes Bachelor of Computer Science, Akwa Ibom State University (2027). AI-training focus includes data types such as Text and labeling workflows including Classification, RLHF, and Prompt + Response Writing (SFT).

Labeling Experience

Self-Directed Prompt Authoring and Adversarial Testing Training and Practice

OtherTextPrompt + Response Writing (SFT)

Authored self-directed prompt datasets to support AI training pipelines across creative, factual, coding, and reasoning categories. Generated adversarial prompts targeting common model failure modes such as hallucination, refusal errors, and reasoning breakdowns. Graded prompt quality using criteria including clarity, specificity, difficulty calibration, and domain coverage to improve training data reliability. • Prompt authoring across multiple categories • Adversarial prompt creation for stress testing • Prompt quality grading via rubrics • Coverage and difficulty calibration of prompt sets

2024 - Present

Self-Directed RLHF Response Evaluation Training and Practice

OtherTextRLHF

Performed self-directed RLHF response evaluation by practicing preference data collection and response ranking against standardized dimensions. Assessed pairs of model outputs for helpfulness, harmlessness, honesty, instruction adherence, coherence, and fluency. Verified factual accuracy and maintained internal rubrics for ambiguous and adversarial edge cases to ensure stable preference labels. • Response comparison and ranking • Instruction-following assessment • Factual accuracy verification • Preference data labeling with ambiguity rubrics

2024 - Present

Self-Directed Text Classification and Tagging Training and Practice

OtherTextClassification

Conducted self-directed text annotation training focused on binary labeling and multi-label classification workflows across large batches. Applied intent detection, sentiment labeling, and toxicity review using detailed rubrics to reduce label drift and maintain consistency across ambiguous cases. Used common annotation task formats including side-by-side comparisons and Likert-scale quality scoring for structured judgments. • Text classification and tagging • Intent, sentiment, and toxicity labeling • Binary and multi-class/multi-label rubric application • Consistency checks to minimize inter-annotator disagreement

2024 - Present

Education

Work History