Senior Python Engineer for AI
Budget: $750 – $1,500 USD
I need a seasoned Python developer who can move fast and set up a robust, Docker-based environment for AI work. Your first mission is to design and code a clean training pipeline for tabular data that I can spin up in a container, run end-to-end tests on, and then extend to large-language-model (LLM) evaluation later.
Key points
• Core task: model training (not just preprocessing or evaluation).
• Data type: tabular; expect CSVs in the tens of millions of rows.
• Tooling: everything must run inside Docker. I’m framework-agnostic right now—if you can show why pure-Python, TensorFlow, PyTorch, or Scikit-Learn is the smartest choice for this dataset, I’m listening.
• Performance and reproducibility matter more than flashy dashboards.
What I’d like to see delivered in the first milestone
1. A Dockerfile that installs all dependencies, exposes clear entry points, and can be built on a vanilla Ubuntu host.
2. Modular Python code for loading data, feature engineering, model definition, training, and serialization.
3. A one-command script (e.g., make train or python train.py) that kicks off the full training cycle and logs metrics to stdout and a local file.
4. Brief README with setup, run instructions, and a note on how the design can later plug into an LLM evaluator.
If we click, this will turn into an ongoing engagement to refine models, add automated evaluation, and productionise the entire pipeline. Let’s talk if you can show recent, hands-on experience with Dockerised AI projects and have the bandwidth to start right away.
Key points
• Core task: model training (not just preprocessing or evaluation).
• Data type: tabular; expect CSVs in the tens of millions of rows.
• Tooling: everything must run inside Docker. I’m framework-agnostic right now—if you can show why pure-Python, TensorFlow, PyTorch, or Scikit-Learn is the smartest choice for this dataset, I’m listening.
• Performance and reproducibility matter more than flashy dashboards.
What I’d like to see delivered in the first milestone
1. A Dockerfile that installs all dependencies, exposes clear entry points, and can be built on a vanilla Ubuntu host.
2. Modular Python code for loading data, feature engineering, model definition, training, and serialization.
3. A one-command script (e.g., make train or python train.py) that kicks off the full training cycle and logs metrics to stdout and a local file.
4. Brief README with setup, run instructions, and a note on how the design can later plug into an LLM evaluator.
If we click, this will turn into an ongoing engagement to refine models, add automated evaluation, and productionise the entire pipeline. Let’s talk if you can show recent, hands-on experience with Dockerised AI projects and have the bandwidth to start right away.