Learner Profiles

Profile 1: Alex, Graduate Student in Molecular Biology

Background: 2 years of research experience, basic Linux skills, uses R frequently

Current Challenge: Alex needs to run genomics workflows on Sagehen that require multiple software versions: - R 4.2.3 for statistical analysis (newer packages needed) - BLAST 2.13 for sequence searching - samtools for manipulating sequencing data - Python for preprocessing scripts

Every time Alex logs into the cluster, they manually load modules and install Python packages, hoping the setup is the same each time. Recently, a collaborator tried to run Alex’s analysis script but got different results because conda packages were installed in a different order.

Why This Workshop: - Wants to save time with module collections - Needs to share reproducible environments with lab members - Confused about whether to use R packages vs. conda vs. pip

What They’ll Get: - Ability to save module restore genomics instead of manually reloading - Knowledge of environment.yml for sharing Python environments - Understanding of job script best practices to ensure reproducibility

Profile 2: Jordan, Postdoc in Physics

Background: Computational physicist, very strong Linux/HPC skills, uses quantum chemistry software

Current Challenge: Jordan runs quantum chemistry simulations with GAMESS and Gaussian, followed by Python analysis. They’ve written sophisticated job scripts but still experience occasional issues: - Sometimes GAMESS fails with cryptic errors about library versions - Different job results depending on which module versions load (by default vs. explicitly) - Wants to create reusable, shareable analysis pipelines that colleagues can run identically

Why This Workshop: - Has intermediate HPC knowledge but wants to optimize workflows - Frustrated by environment-related debugging - Wants to understand best practices for complex multi-software pipelines

What They’ll Get: - Deep understanding of module dependencies and version pinning - Knowledge of conda for reproducible Python analysis stacks - Job script patterns that work across different users and times

Profile 3: Sam, Undergraduate in Data Science

Background: One semester of Python experience, no HPC experience, never used Linux before workshop

Current Challenge: Sam needs to use machine learning libraries (TensorFlow, scikit-learn) on Sagehen HPC for a research project but: - Doesn’t understand what “modules” are or why they’re necessary - Installed packages with pip install --user globally and now confused about versions - Script works on their laptop but fails on Sagehen

Why This Workshop: - Complete beginner who needs foundational understanding - Wants to create a single reproducible environment for their project - Needs confidence that their setup is correct

What They’ll Get: - Clear explanation of why environment management is important - Step-by-step conda environment setup for machine learning - Templates for reproducible job scripts - Understanding of when to ask for help ()

Accommodating Different Profiles

This workshop is designed to serve all three profiles simultaneously:

For Alex (intermediate): - Episodes 1-3 review and solidify existing knowledge - Episodes 4-6 address specific gaps (conda, best practices) - Challenge exercises have increasing difficulty

For Jordan (advanced): - Episodes 1-2 are familiar; can be quickly reviewed - Episodes 3-6 provide structure and best practices for complex workflows - Instructor notes and references serve as documentation - Challenge exercises allow deeper exploration

For Sam (beginner): - Episodes 1-2 provide essential foundation - Episode 4 (conda) is most relevant for their Python work - Ample time for questions and clarification - Reference card serves as long-term resource

Timing Flexibilities

For pure beginners (Sam): - Spend extra time on episodes 1-2 - Skip episode 5 (pip/venv) if time is short - Focus on episode 4 (conda) and episode 6 (best practices) - Allow 3 hours instead of 2.5 hours

For intermediate (Alex): - Episodes 1-2 quick review (15 min instead of 40 min) - Full depth on episodes 3-6 (120 min) - Discuss real-world genomics workflows - Allow 2.5 hours as planned

For advanced (Jordan): - Episodes 1-3 brief review (20 min) - Episodes 4-6 with depth and discussion of edge cases (60 min) - Time for discussion about their specific use cases - Use full 2.5 hours for in-depth problem-solving

Pre-Workshop Assessment

To gauge learner profiles, send a survey before the workshop:

  1. How would you describe your Linux experience?
    • Never used it / Some experience / Comfortable / Advanced
  2. Have you used HPC clusters before?
    • No / Yes, but briefly / Yes, regularly / Yes, extensively
  3. Which software do you primarily use?
    • Python / R / Other compiled software / Multiple languages
  4. What’s your main goal for this workshop?
    • Understand modules / Manage Python environments / Reproducibility / Advanced optimization

Based on responses, you can: - Assign reading materials pre-workshop - Create advanced/beginner breakout groups - Prepare specific examples relevant to audience - Adjust pacing and focus