Browse categories
Explore
Fiverr Pro
English
$
USD
I will build computer vision solutions with yolo, CNN, and opencv
Rokibul
About this gig
Struggling to turn raw images or video into a working AI system? I build custom computer vision solutions that actually get deployed, not just notebook demos.
I specialize in object detection, image segmentation, and object tracking using YOLO, CNN architectures, and OpenCV. My work spans fine-tuning models on custom datasets, optimizing accuracy through data augmentation and architecture selection, and packaging results into a usable API not just a model file you don't know what to do with.
Recent work includes fine-tuning YOLOv8 for real-world object detection, building self-supervised segmentation models for medical imaging, and developing brain-inspired object detection systems combining CNN backbones with attention mechanisms.
What you get:
Custom-trained model on your dataset (not a generic pretrained demo)
Clear evaluation metrics (accuracy, mAP, F1) so you know it actually works
Clean, documented code Python, PyTorch, OpenCV
Optional FastAPI deployment so your model is usable, not just trained
I work end-to-end: data preprocessing, model training, evaluation, and deployment. Whether you need a quick prototype or a production-ready pipeline.
Get to know Rokibul
Rokibul
AI,ML Engineer
- FromBangladesh
- Member sinceJul 2026
- Avg. response time1 hour
Languages
English, Hindi
AI/ML Engineer specializing in computer vision, LLM/VLM fine-tuning, and RAG systems. Experience fine-tuning OpenVLA, SmolVLA, and BanglishBERT; building CNN-based segmentation/detection pipelines; and developing agentic AI workflows with LangGraph and MCP. Comfortable across the full pipeline — data processing, training, evaluation, deployment (FastAPI/Django). Native Bangla speaker, strong edge in code-mixed NLP. CSE graduate (AI/ML concentration), researching robotic scene understanding. Reliable communication, clean code, and real problem-solving.
My Portfolio
FAQ
What kind of dataset do I need to provide?
Any labeled or unlabeled images/video relevant to your use case. If your data isn't labeled yet, I can help with annotation guidance or use pretrained models as a starting point — just share what you have and I'll advise on the best approach.
Can you work with a small dataset?
Yes. I use techniques like data augmentation, transfer learning, and fine-tuning pretrained models (YOLOv8, CNN backbones) to get strong results even with limited data — this is common in real-world CV projects.
Do I get the source code and trained model?
Yes, all tiers include the trained model file and clean, documented code. Higher tiers also include deployment scripts and a technical report explaining the architecture and results.
What if my use case doesn't fit the standard packages?
No problem — message me with your requirements before ordering. I'll scope a custom offer if your project needs something outside the listed packages (e.g. video-based tracking, real-time inference, or multi-model pipelines).
How do you ensure the model actually works, not just trains?
Every delivery includes an evaluation report with real metrics (accuracy, mAP, F1-score, confusion matrix where relevant) so you can verify performance before deploying — not just take my word for it.
Can you deploy the model for me?
Yes, Standard and Premium tiers include a FastAPI deployment so the model is callable via an API endpoint, not just a raw file. Let me know your hosting preference, and I'll advise on integration.
