Md. Roushan Alam

Key Skills

Software

AWS SageMaker

Appen

CrowdSource

Data Annotation Tech

Google Cloud Vertex AI

OneForma

Scale AI

SuperAnnotate

Toloka

V7 Labs

Internal/Proprietary Tooling

Don't disclose

Top Subject Matter

No subject matter listed

Top Data Types

3D Sensor

Audio

Computer Code Programming

Document

Geospatial Tiled Imagery

Image

Text

Video

Top Label Types

Action Recognition

Audio Recording

Computer Programming Coding

Evaluation Rating

Fine Tuning

Prompt Response Writing SFT

Text Generation

Text Summarization

Transcription

Freelancer Overview

AI Trainer and LLM Evaluator with experience in AI response grading, prompt engineering, and structured model feedback. Skilled in assessing logical reasoning, factual accuracy, instruction adherence, bias risk, and response completeness. Experienced in improving AI outputs through clear, systematic evaluation frameworks. Strong analytical background with cross-domain expertise in technology, education, and business strategy.

Labeling Experience

ntent Classification and Function Calling Optimization

Don T DiscloseDocumentEntity Ner ClassificationQuestion Answering

Annotated user prompts for intent classification and API routing accuracy. Evaluated structured outputs for correct function invocation, parameter extraction, and response formatting. Classified task types and edge cases to improve automation workflows and fine tuning pipelines for tool enabled language models.

Annotated user prompts for intent classification and API routing accuracy. Evaluated structured outputs for correct function invocation, parameter extraction, and response formatting. Classified task types and edge cases to improve automation workflows and fine tuning pipelines for tool enabled language models.

2025 - 2025

Speech Data Collection and Conversational AI Evaluation

AppenAudioClassificationQuestion Answering

Recorded structured speech datasets for conversational AI training. Reviewed and classified audio quality, transcription accuracy, pronunciation clarity, and intent recognition performance. Evaluated AI generated spoken responses for clarity, alignment, and correctness in conversational contexts.

Recorded structured speech datasets for conversational AI training. Reviewed and classified audio quality, transcription accuracy, pronunciation clarity, and intent recognition performance. Evaluated AI generated spoken responses for clarity, alignment, and correctness in conversational contexts.

2024 - 2025

AI Code Generation Validation and Programming Assessment

Internal Proprietary ToolingComputer Code ProgrammingClassificationRed Teaming

Reviewed AI generated code for correctness, security vulnerabilities, and logical accuracy. Evaluated syntax, edge case handling, and instruction compliance. Conducted adversarial testing to identify unsafe or inefficient code patterns. Classified outputs based on functionality level and reliability before feedback submission for model improvement.

Reviewed AI generated code for correctness, security vulnerabilities, and logical accuracy. Evaluated syntax, edge case handling, and instruction compliance. Conducted adversarial testing to identify unsafe or inefficient code patterns. Classified outputs based on functionality level and reliability before feedback submission for model improvement.

2024 - 2025

LLM Alignment, RLHF and Instruction Compliance Evaluation

Internal Proprietary ToolingTextClassificationQuestion Answering

Evaluated AI generated responses across analytical, factual, and multi-step reasoning tasks. Conducted pairwise ranking under RLHF workflows, classified outputs based on instruction adherence and quality levels, and assessed question answering accuracy. Provided structured reasoning feedback focusing on hallucination detection, bias identification, logical consistency, and completeness to improve model alignment and response reliability.

Evaluated AI generated responses across analytical, factual, and multi-step reasoning tasks. Conducted pairwise ranking under RLHF workflows, classified outputs based on instruction adherence and quality levels, and assessed question answering accuracy. Provided structured reasoning feedback focusing on hallucination detection, bias identification, logical consistency, and completeness to improve model alignment and response reliability.

2024 - 2025

Education

Work History