Browse categories
Explore
Fiverr Pro
English
$
USD
I will build ocr and computer vision solutions with python and opencv
Harshal D
About this gig
Need OCR, image classification, object detection, or a custom computer vision pipeline? I build production-grade CV systems that actually work outside a Jupyter notebook.
WHAT I DELIVER:
Document OCR (invoices, receipts, forms, handwritten text)
Image classification with custom CNNs
Object detection and tracking (YOLO,detectron)
Multimodal pipelines (OCR + NLP + LLM post-processing)
Indian-language OCR (Hindi, Marathi, Indic scripts)
Production deployment with Docker and REST APIs
WHAT YOU GET:
- Clean, documented Python code you own
- Preprocessing tuned for your image quality
- Accuracy validation on your real data
- Docker container ready to deploy(Premium)
- REST API endpoints (Standard and Premium)
TECH STACK:
OpenCV, Tesseract, EasyOCR, PaddleOCR
TensorFlow, PyTorch, Keras
YOLO, Detectron2,custom CNNs
FastAPI, Flask,Docker
MY CV TRACK RECORD:
- Shipped Whisper + OCR video pipeline at Sambhav AI (50% faster, deployed on Kubernetes)
- Published CNN research in IJCNIS (Skin cancer classifier, 80%+ TPR)
- Breast cancer prediction model (97% accuracy on 10K+ records)
- Built OCR-powered POS invoice parser (ISKCON,ShopMind)
- GitHub: github.com/harshaldonarkar
Message me
Get to know Harshal D
Harshal D
AI Engineer: RAG Pipelines and LLM Integration Expert
- FromIndia
- Member sinceApr 2022
Languages
Hindi, Marathi, English
AI Engineer with 2 published research papers (IJCNIS, IRJET) and production deployments. Previously at Sambhav AI where I built Whisper+OCR pipelines and multimodal AI — delivered 50% performance gains. I specialize in RAG systems with LangChain, LLM integration (GPT-4, Claude, Gemini), computer vision with OpenCV, OCR pipelines, and production ML deployment with Docker. I ship working code, not prototypes. Portfolio: harshaldonarkar.github.io | GitHub: harshaldonarkar
My Portfolio
Other AI Development Services I Offer
FAQ
What image quality do I need?
I'll recommend preprocessing; most real-world images (phone photos, scans, screenshots) work with the right pipeline. Share samples and I'll tell you upfront.
Can you handle handwritten text?
Yes — EasyOCR or custom fine-tuning depending on volume and handwriting style. Share samples for an accuracy estimate.
What about Indian-language OCR?
Yes — Hindi, Marathi, and other Indic scripts are supported. Available as a Premium feature or as a paid extra on Basic/Standard.
Can you combine OCR with LLM post-processing?
Absolutely — this is one of my strengths. Extract text → understand context → structure output. Great for invoices, forms, and unstructured documents.
Do you deploy the model or just deliver code?
Basic and Standard deliver code + REST API. Premium includes Docker deployment, ready to run on your server or cloud.
Can you train a custom model for my dataset?
Yes — custom CNN training is included in Premium, or available as an extra. I'll need labeled training data from you.
What accuracy can I expect?
Depends heavily on your data. For clean printed text OCR, 95%+ is typical. For handwritten or degraded images, we validate on samples first.
Do you handle real-time video processing?
Yes — object detection and tracking on video streams is available as a paid extra. Happy to discuss frame rate and latency requirements.
