Looking for a job

Mushfiqur Rahman

@mushfiqurrahman1

Message

I evaluate Bengali and English AI outputs for accuracy, safety, fluency, and instruction adherence.

Bangladesh

Message

What I'm looking for

I'm looking for remote, asynchronous AI language and evaluation work where I can apply Bengali-English fluency, rigorous quality standards, factual verification, and structured feedback to improve model performance.

I'm a Bengali-English AI language evaluator who assesses LLM outputs for factual accuracy, contextual relevance, safety, tone, linguistic fluency, and instruction following.

Across remote AI training projects, I've compared parallel model responses, identified hallucinations and logical flaws, created Bengali and English test prompts, and written evidence-backed scoring rationales. I've also transcribed and annotated audio and video datasets while maintaining QA/QC standards in asynchronous, high-throughput workflows.

My B.Pharm studies and debate adjudication experience strengthen my scientific reasoning, source verification, and ability to deliver structured, objective feedback in both Bengali and English.

Experience

Work history, roles, and key accomplishments

Debate Adjudicator & Competitive Debater

Gono Bishwabidyalay Debating Society

Analyzed complex arguments and evidence credibility in bilingual debate formats. Served as adjudicator delivering structured qualitative evaluations.

Argument Analysis Evidence Evaluation Bilingual Communication Qualitative Assessment

AI Data & Language Specialist

Remote / Freelance

Executed multi-modal AI data tasks including text evaluation, transcription, data labeling, and cross-lingual validation. Assessed model fluency and factual consistency across Bengali and English.

Text Evaluation Speech to text transcription Data Labeling Cross Language Validation LLM Benchmarking Prompt Testing

Bengali AI Trainer & Language Evaluator

AI Training Projects

Evaluated AI-generated responses for accuracy, relevance, tone, and instruction adherence. Authored evidence-backed rationale and designed prompts to probe model boundaries.