Freelance Machine Learning Engineer to design and develop computational STEM problems requiring Python programming and non-trivial reasoning for AI testing and evaluation at Mindrift. This is a project-based role with flexible weekly hours, not permanent employment.
What You'll Do
- Design original computational STEM problems that simulate real scientific workflows
- Create problems that require Python programming to solve
- Ensure problems are computationally intensive and cannot be solved manually within reasonable timeframes (days/weeks)
- Develop problems requiring non-trivial reasoning chains and creative problem-solving approaches
- Verify solutions using Python with standard libraries (Numpy, Pandas, Scipy, scikit-learn)
- Document problem statements clearly and provide verified correct answers
What We're Looking For
- 5+ years of hands-on machine learning experience with proven business impact
- Portfolio of completed projects and publications showcasing real-world problem-solving
- Expert Python programming for data science (pandas, numpy, scipy, scikit-learn, statsmodels)
- Expert statistical analysis and machine learning - deep understanding of algorithms, methods, and their practical applications
- Expert with SQL and database operations for data manipulation and analysis
- Experience with GenAI technologies (LLMs, RAG, prompt engineering, vector databases)
- Understanding of MLOps practices and model deployment workflows
- Knowledge of modern frameworks (TensorFlow, PyTorch, LangChain)
- Strong written English (C1+)
Technical Stack
Python, Numpy, Pandas, Scipy, scikit-learn, statsmodels, SQL, TensorFlow, PyTorch, LangChain, LLMs, RAG, prompt engineering, vector databases
Benefits & Compensation
- Project-based work with leading tech companies
- Opportunity to work on AI testing, evaluating, and improving AI systems
Compensation: $58 per hour equivalent. Compensation varies across projects depending on scope, complexity, and required expertise.
Work Mode
Tasks estimated at 10–20 hours per week during active phases; not a guaranteed workload. Fully remote with global availability.