ML (computer vision ) engineering project

Budget: 150 USD Deadline: 3 days

Hello!

Your project is exactly the kind of engineering challenge I enjoy—combining computer vision, machine learning, and robust data engineering into a production-ready system rather than focusing only on model training.

I have experience building end-to-end ML pipelines using Python and PyTorch, from data ingestion and preprocessing through training, evaluation, deployment, and API integration. I believe successful computer vision projects depend as much on data quality, reproducible pipelines, and rigorous evaluation as they do on model architecture.

My approach would include:

* Building a scalable pipeline for image ingestion, preprocessing, feature extraction, training, and inference.
* Implementing classical computer vision techniques alongside modern deep learning models to establish reliable baselines before increasing model complexity.
* Training and fine-tuning vision models for detection, anomaly detection, and representation learning where appropriate.
* Working with PostgreSQL, feature storage, experiment tracking, and reproducible dataset versioning.
* Designing evaluation methodologies that measure real-world performance rather than relying solely on benchmark metrics.
* Developing maintainable APIs and inference services for production deployment.

I am comfortable working with technologies such as Python, PyTorch, OpenCV, scikit-image, NumPy, SciPy, scikit-learn, FastAPI, PostgreSQL, Redis, MLflow, Docker, and modern MLOps practices. Your proposed stack aligns well with how I typically structure production ML systems.

I also appreciate your emphasis on pragmatic engineering. My preference is to validate assumptions with simple statistical or classical CV approaches first, establish measurable baselines, and then introduce more sophisticated models only when they provide a clear improvement.

I would be happy to discuss your dataset, the characteristics of the industrial imagery, annotation quality, evaluation criteria, and deployment environment. Once I understand the workflow, I can propose an implementation plan that balances accuracy, reliability, and maintainability.

I look forward to the opportunity to collaborate on this project. Expand

Budget: 700 USD Deadline: 5 days

Hello! I am Serhiy, I have 8 years of experience in IT across various technologies: web/mobile, backend, API, integrations, automation, and data work.
I see the main task as follows: A project in ML engineering (computer vision). This project sits at the intersection of computer vision and classical data engineering. We are building a...

Skills for the task:
- backend/API and integrations
- AI/ML, RAG, and automation
- bots, parsing, and data work

Portfolio: https://software.campstudio.net/#showcases

Budget: 150 USD Deadline: 5 days

This project is exactly the case where a classic CV and deep learning should work together, not compete. I have seen many ML engineers rush into training models, ignoring preprocessing and baselines, and then wonder why the metrics are misleading. Therefore, my approach is: first, a thorough data analysis, building simple statistical baselines, and only then fine-tuning modern architectures. I have over 5 years of experience with Python and PyTorch, have worked with 16-bit scientific images, so I understand the specifics of processing such data. I can independently build a pipeline: from ingestion in PostgreSQL (with pgvector for vector searches) to serving via FastAPI. Of course, with validation at each stage—no “magic” numbers. If needed, I can integrate DVC for data versioning and Prefect for orchestration. I estimate the work at $1200-1800 depending on the volume of data and complexity of the models. The timeline is 21 days for the MVP, then iteratively.

Budget: 92 USD Deadline: 5 days

I'd treat this as two tracks meeting in Postgres: image features from scans versus structured records as ground truth, mismatches flagged for review.

First step: preprocessing 16-bit imagery (scikit-image/OpenCV), then a simple statistical baseline for anomalies before reaching for a PyTorch model.

Could show a working pipeline slice in about 5 days, once you confirm if raw format is DICOS.

Budget: 1000 USD Deadline: 12 days

Hello, I can help with the computer vision and data engineering part: Python, PyTorch, image processing, structured data matching, and inconsistency detection.
I would start by clarifying the scan format, record schema, and validation rules, then build a pipeline for preprocessing, feature/model inference, matching, and human-review flags.
For the backend/data layer, I can use Python with FastAPI or a batch-processing service, depending on how the system should be operated.
I have relevant experience with Python backends and production-oriented data workflows, including backend work for ZEM Center: https://zem.center/
Ready to discuss the current dataset structure and propose a practical implementation plan.
git: github.com/onyx144

Budget: 82 USD Deadline: 4 days

Task: verification of scans with the database, discrepancy flags. Preprocessing 16-bit, statistical baseline, then detection model.

Postgres with pgvector, MLflow pipeline.

Scans in DICOS or another format? Result approximately in 4 days.