Browse categories
Explore
Fiverr Pro
English
$
USD
I will fine tune openai whisper on your custom dataset
M. Sabtain Khan
About this gig
Looking for a custom Speech-to-Text (ASR) solution using OpenAI Whisper?
I will help you fine-tune OpenAI Whisper on your custom dataset to improve transcription accuracy for your language, accent, or domain. Whether you're working on academic research, a business application, or an AI project, I'll deliver a clean, well-documented, and reliable solution.
What I Can Do
- Fine-tune OpenAI Whisper using your custom dataset
- Prepare and preprocess audio datasets
- Train the model using LoRA (PEFT)
- Evaluate model performance using Word Error Rate (WER)
- Provide inference scripts for transcription
- Integrate a FastAPI inference API (Premium)
- Dockerize the project for easy setup (Premium)
- Deliver clean, documented Python source code
What You'll Receive
- Fine-tuned Whisper model
- Training and inference scripts
- Source code
- Evaluation results
- Documentation
Buyer Requirements
Please provide:
- Your speech/audio dataset (or dataset link)
- Matching transcripts
- Access to GPU resources (Google Colab Pro, Kaggle, RunPod, AWS, or your own machine) for model training
- Project requirements and expected output
Please contact me before placing an order so we can discuss your project.
Get to know M. Sabtain Khan
M. Sabtain Khan
Machine Learning Engineer
- FromPakistan
- Member sinceFeb 2026
- Avg. response time1 hour
Languages
Urdu, English
I'm a Machine Learning Engineer with hands-on experience in Python, Machine Learning, Natural Language Processing (NLP), and Speech Recognition. My primary project involved fine-tuning OpenAI Whisper for Pashto Speech-to-Text using a custom dataset, where I worked on data preprocessing, model training, evaluation, and performance improvement.
I enjoy building reliable AI solutions, writing clean and maintainable Python code, and solving real-world problems with machine learning. I'm always eager to learn new technologies, collaborate with clients, and deliver quality work on time.
My Portfolio
FAQ
Q: Do I need to provide a dataset?
A: Yes. You should provide your audio dataset (or a downloadable dataset link) along with the corresponding transcripts. If you don't have a dataset, feel free to contact me before placing an order.
Q: Do I need to provide GPU resources?
A: Yes. Whisper fine-tuning requires significant computational resources. Please provide access to a GPU environment such as Google Colab Pro, Kaggle, RunPod, AWS, or your own machine.
Q: Which Whisper models can you fine-tune?
A: I can work with Whisper Tiny, Base, Small, Medium, and Large models, depending on your project requirements and available GPU resources.
Q: Will I receive the source code?
A: Yes. All packages include clean, well-documented Python source code. Depending on the package, you'll also receive training scripts, inference scripts, and project documentation.
Q: Can you improve an existing Whisper model?
A: Yes. If you already have a trained Whisper checkpoint, I can continue fine-tuning it on your custom dataset or adapt it for a new language or domain.
Q: Can you deploy the model?
A: The Premium package includes Docker support and a FastAPI inference API. Full cloud deployment (AWS, Azure, GCP, etc.) is not included unless discussed separately.
Q: How do I know which package is right for me?
A: Send me a message before ordering. I'll review your dataset, project requirements, and available resources, then recommend the most suitable package.
