For employers

Hire this AI Trainer

Sign in or create an account to invite AI Trainers to your job.

Invite to Job
Muhammad Ahsan

Muhammad Ahsan

AI Engineer - Smart, Cost effective AI Solutions for Startups

PAKISTAN flag
Karachi, Pakistan
$35.00/hrIntermediateInternal Proprietary ToolingDon T Disclose

Key Skills

Software

Internal/Proprietary Tooling
Don't disclose

Top Subject Matter

No subject matter listed

Top Data Types

ImageImage
DocumentDocument

Top Label Types

Audio Recording
Data Collection
Evaluation Rating
Fine Tuning
Object Detection
Prompt Response Writing SFT
Question Answering
RLHF
Transcription

Freelancer Overview

I am an AI Engineer with hands-on experience in developing and deploying LLM-powered applications, computer vision systems, and NLP solutions, with a strong focus on data labeling, annotation, and curation for training robust AI models. My work includes building domain-specific chatbots, fine-tuning large language models, and creating custom datasets for tasks like sentiment analysis, topic classification, and CAPTCHA recognition, where I have personally managed data collection, self-labeling, and preprocessing to ensure high-quality training inputs. I am skilled in Python, PyTorch, TensorFlow, HuggingFace, OpenCV, and vector databases, and have utilized tools like LangChain and Unsloth for efficient model adaptation. My projects span domains such as petroleum industry document processing, accessibility tools for the visually impaired, and real-time data extraction, all relying on well-structured and accurately labeled datasets to achieve production-ready results. I am passionate about building scalable, cost-efficient AI systems and continuously optimizing data workflows to improve model performance.

IntermediateEnglishUrdu

Labeling Experience

LLM Finetuning on custom documents for specialized responses

Don T DiscloseDocumentQuestion AnsweringRLHF
I contributed to a dataset creation effort for fine-tuning large language models (LLMs) by producing high-quality instruction–response pairs tailored to a specific enterprise domain. The scope involved identifying relevant operational scenarios, drafting clear end-user instructions, generating correct and domain-aligned responses, and formatting outputs into a structured JSON format suitable for supervised fine-tuning. The data labeling task focused on instruction engineering, response authoring, and quality control to ensure factual correctness, consistency in style, and alignment with the intended model behavior. Internal proprietary guidelines were followed for tone, format, and reasoning depth. Quality was ensured through multi-pass review, rejection of ambiguous prompts, normalization of labels.

I contributed to a dataset creation effort for fine-tuning large language models (LLMs) by producing high-quality instruction–response pairs tailored to a specific enterprise domain. The scope involved identifying relevant operational scenarios, drafting clear end-user instructions, generating correct and domain-aligned responses, and formatting outputs into a structured JSON format suitable for supervised fine-tuning. The data labeling task focused on instruction engineering, response authoring, and quality control to ensure factual correctness, consistency in style, and alignment with the intended model behavior. Internal proprietary guidelines were followed for tone, format, and reasoning depth. Quality was ensured through multi-pass review, rejection of ambiguous prompts, normalization of labels.

2024 - 2025

Custom model training for Captcha recognition

Internal Proprietary ToolingImageTranscription
I developed a Captcha Recognition model for a client, for a website that contained alphanumeric CATCHA images different form the Google ones. The scope involved scraping 3000 raw CAPTCHA images form the website, preparing for training, and performing manual text transcription to create labels. The data labeling task required identifying and transcribing 6–9 character strings for each image. Quality was ensured through double-pass verification, consistency checks on character sets, and enforcing annotation guidelines for ambiguous characters.

I developed a Captcha Recognition model for a client, for a website that contained alphanumeric CATCHA images different form the Google ones. The scope involved scraping 3000 raw CAPTCHA images form the website, preparing for training, and performing manual text transcription to create labels. The data labeling task required identifying and transcribing 6–9 character strings for each image. Quality was ensured through double-pass verification, consistency checks on character sets, and enforcing annotation guidelines for ambiguous characters.

2024 - 2024

Education

N

NED University of Engineering and Technology

Bachelor of Science, Computer Science

Bachelor of Science
2022 - 2025
A

Adamjee Government Science College

High School Diploma, Pre-Engineering

High School Diploma
2020 - 2022

Work History

F

Freelance

Freelance Developer

Karachi
2024 - Present
D

Devairo

AI/ML Developer

Karachi
2024 - 2025