Building ML in the Dark: A Survival Guide for the Solo Practitioner

The constraints you're working under — no compute, bad data, and no engineering support — are common challenges for solo ML practitioners. Here’s how to navigate them effectively:

1. Start with Data Auditing: Spend time understanding your dataset before diving into model training. Use scripts like the one provided to check for duplicates, variance in labeling, and other issues that could skew results.

2. Simplify Model Selection: Opt for models that are interpretable and require minimal tuning, such as logistic regression or gradient-boosted trees (XGBoost, LightGBM). These models often provide sufficient performance without the complexity of more advanced techniques.

3. Focus on Simplicity in Deployment:

   • Use FastAPI with Docker for small models.
   • For tasks requiring GPUs, consider using services like Modal to avoid infrastructure management.
   • Batch scoring can be a viable option if real-time latency isn't critical.

4. Refuse Ambiguous Requests: Clearly articulate why certain requests aren’t feasible without proper context or data. Offer concrete alternatives such as rule-based systems or lightweight annotation workflows.

5. Prioritize Quick Wins: Deploy models that are 70-80% effective and use them to gather feedback for

Read the full article at Towards AI - Medium

---

Want to create content about this topic? Use Nemati AI tools to generate articles, social posts, and more.