Parbhej S.

Key Skills

Software

OneForma

Other

Top Subject Matter

No subject matter listed

Top Data Types

Document

Image

Text

Top Task Types

Data Collection

Evaluation/Rating

Prompt + Response Writing (SFT)

RLHF

Translation/Localization

Freelancer Overview

I am an AI Data Annotator specializing in LLM training, RLHF workflows, and multimodal evaluation across text, audio, image/OCR, and translation tasks. At RWS, I work on high-impact AI training pipelines where I maintain 95%+ DQA accuracy while performing single-review annotation in platforms such as GALA/HALO SRT, Multimango, and TryRating-type systems. My responsibilities include ranking LLM outputs, validating model predictions, improving search relevance, and performing high-fidelity transcription and OCR review. I have contributed to projects involving web-crawled data auditing, spam detection, HTML/encoding cleanup, translation quality evaluation (MTPE), and linguistic safety review. My background also includes prompt engineering, generative AI, and hands-on data analysis using tools like Excel, Power BI, and Google Analytics. With strong attention to detail and experience training multilingual and multimodal systems, I bring reliable, scalable data annotation expertise for next-generation AI models.

Labeling Experience

Search Relevance Evaluation & Query Quality Rating (MilkyWay/Lightspeed)

OneFormaTextText SummarizationRLHF

Evaluated search results for accuracy, usefulness, safety, and intent relevance using OneForma’s MilkyWay/Lightspeed platform. Judged the alignment between user queries and returned results, applied detailed rating guidelines, flagged unsafe or low-quality content, and improved ranking algorithms through consistent, high-precision evaluations. Responsibilities also included analyzing user intent, scoring page quality, and providing granular feedback that enhanced search engine performance for global clients. Maintained 95%+ accuracy across tasks while adapting quickly to guideline updates and new rating categories.

2025

Translation Quality Evaluation & Linguistic Safety Review (Project Ocean)

OtherTextTranslation/LocalizationRLHF

Reviewed machine translations for linguistic accuracy, tone, cultural relevance, and safety compliance. Applied MTPE techniques to identify mistranslations, stylistic issues, and problematic content. Provided final quality ratings and corrective suggestions to improve MT models. Contributed to safety filtering pipelines by identifying harmful or unsafe outputs.

2025 - 2025

Multimodal Transcription & OCR Output Validation (Project Diamond)

OtherAudioRLHFEvaluation/Rating

Evaluated model-generated audio transcriptions and OCR outputs to select the most accurate result for training multilingual speech and vision models. Compared AI predictions against provided media, ranked outputs, and corrected errors to build high-quality ground truth datasets. Ensured adherence to linguistic accuracy, formatting rules, and quality standards while consistently maintaining 95%+ DQA accuracy.

2025 - 2025

Web-Crawled Text Quality Audit & LLM Training Support (Project Avocado)

OtherTextRLHFEvaluation/Rating

Reviewed large-scale web-crawled datasets to ensure high-quality text inputs for LLM training. Tasks included removing spam, correcting HTML and encoding issues, filtering harmful or low-quality content, and applying annotation guidelines to classify data appropriately. Contributed to improving the training corpus by applying consistent quality standards and providing feedback on problematic sources. Maintained 95%+ DQA accuracy across all assignments.

2025 - 2025

Education

Work History