Md. Kawser A.

Key Skills

Software

Scale AI

Top Subject Matter

AI model alignment (RLHF)

prompt engineering

and technical QA for AI systems

Top Data Types

Document

Image

Text

Top Task Types

RLHF

Segmentation

Freelancer Overview

Independent Contractor · Freelance (Scale AI / Outlier) — RLHF feedback and technical review. Brings 2+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include Scale AI and OpenCV. Education includes Bachelor of Science, BRAC University (2026) and Higher Secondary Certificate, Bangladesh Noubahini School And College (2021). AI-training focus includes data types such as Computer Code, Programming, and Document and labeling workflows including RLHF and Segmentation.

Labeling Experience

Independent Contractor (Technical Code Reviewer) - Scale AI

Scale AIRLHF

You worked as an independent technical code reviewer on an AI RLHF (reinforcement learning from human feedback) project. You reviewed and debugged AI-generated code across Python, JavaScript, and C to identify logic errors, edge cases, and security flaws. You provided prompt engineering and RLHF feedback to support responsible model alignment and checked technical outputs for syntax and real user functionality needs. • Reviewed and debugged AI-generated code in Python, JavaScript, and C • Provided RLHF feedback to support model alignment and accuracy • Applied prompt engineering to test model behavior under varied inputs • Performed QA for syntax and functional correctness against user needs

2025 - Present

Independent Contractor · Freelance (Scale AI / Outlier) — RLHF feedback and technical review

Scale AIRLHF

Worked as a technical code reviewer for an RLHF (reinforcement learning from human feedback) alignment workflow under a contract with Scale AI (Outlier). Reviewed and debugged AI-generated code to ensure model behavior stayed correct, safe, and ethically aligned. Provided RLHF feedback that directly contributed to model alignment and output accuracy. • Performed QA on technical outputs by checking syntax and functional behavior against user needs. • Used prompt engineering to test and guide model responses under different inputs and conditions. • Identified logic errors, edge cases, and security flaws during review. • Ensured outputs remained within legal and ethical boundaries throughout the process.

2025 - Present

Research project — Deep Learning handwritten text recognition using a lightweight two-stage OCR model

DocumentSegmentation

Built an end-to-end bilingual OCR deep learning pipeline for handwritten Bengali and English documents. Implemented a YOLOv8n segmentation-based detector to localize words and extract bounding boxes for downstream recognition. Engineered a hybrid recognition model using spatial transformer networks, MobileNetV3-Small feature extraction, and transformer encoders for sequence modeling. • Applied CTC loss with grapheme-level tokenization for sequence prediction. • Used Unicode NFC normalization to correctly handle complex Bengali conjuncts. • Designed word-level localization and bounding-box extraction using segmentation outputs. • Coordinated the full OCR pipeline stages from detection through recognition for document inputs.

2025 - 2026

Education

Work History