Many scientists in the biopharma industry think they’re underusing Large Language Models for their science. Is it just a chat interface or an email writer? Something is missing.

My colleague Sonia Timberlake and I adapted best practices from AI superteams into workshops designed for life science professionals. In these workshops, we provide hands-on case studies for participants to experience the power (and edges) of the best LLMs using examples close to their work. This includes writing research summaries, analyzing data, and building pipelines.

Here’s the list of what we offer, either as a drop-in class using public data or customized to our clients’ internal systems.

Agentic coding for computational biology

Workshop name: Agentic coding for Computational Biology

Problem statement

Agentic coding is increasing the productivity of software developers. What agentic coding practices can be applied to computational biology?

We have adapted software development workshops into the “Scientific Specification-Driven Development” method, which employs spec design with agentic validation loops that respond to data.

See the examples of the final projects here on the class website.

Learning goals

• Adopt agentic coding best practices from software dev
• Spec-driven development with LLM critique
• Adapt AI systems with domain expertise
• Provide the right context, skills, and tools

Example Case Study

Use a foundation model to process single cell genomics data and answer drug development questions. Develop “skills” to navigate public data sets and employ MCPs to test hypotheses.

Walk away with an interactive dashboard comparing public data to predictions of cell types using a foundation model, developed in under 3 hours.

Audience

Computational biologists in biotech or pharma

Workflow management for computational biology

Workshop name: Use AI to turn your scripts into a pipeline

Problem statement

Computational biologists often maintain analysis scripts with the intention of turning them into a pipeline, but don’t have the time to learn a pipeline language or containerization patterns. Now, LLMs can turn functioning scientific scripts into a pipeline, complete with environment and testing suite.

Learning goals

• Have AI write a pipeline (e.g. with Snakemake/WDL/etc) from analysis scripts and common computational biology tools
• Architectural basics for pipelines
• Dockerization and reproducible environments
• Writing and running tests

Example Case Study

Multistep genomic analysis from downloading data, pre-processing, and aligning to references to creating an interactive report.

Audience

Computational biologists in biopharma (also applies to academics)

AI for knowledge work in biopharma

Workshop name: Best Practices for AI for knowledge work in biopharma

Problem statement

AI can do more than write emails; it can be a trusted scientific collaborator across research and development. Participants develop an intuition for the strengths and limitations of the latest models so that LLMs can be safely integrated into scientific workflows.

Learning goals

• Give the LLM the right autonomy, tools, and context to be an effective co-worker
• Literature review with LLMs
• Develop data-driven intuition for current strengths and weaknesses of models
• Summarize data into effective communications

Example Case study

Create a research proposal relevant to your work that summarizes existing data and identifies next steps, with multiple review points to limit hallucinations.

Audience

All scientists, from RAs to VPs

AI upskilling for IT professionals

Workshop name: Use LLMs to Deploy Secure Cloud Infrastructure for Enterprise Life Sciences

Problem statement

If you’re an IT leader in the life sciences, you probably rely on a suite of MSPs to meet the fast-moving needs of your scientific customers. Besides being a budget-breaker, this can be a headache to manage. To what extent can we use AI to bring these capabilities in-house, reducing cost and increasing compliance?

Learning goals

• Learn access patterns that scientists need from cloud resources
• Learn how cloud changes security paradigms
• Use AI to Terraform a cloud environment (i.e. with built-in controls)
• Create a configuration release and control process with human and AI checks
• Create a corporate knowledge base with a help agent

Example Case study

• A Terraform environment for a practical cloud application (i.e. “a scientist needs to deploy this” and this is the way to securely do it)
• Auto-generated cloud diagram with security checks
• Change management SOPs and skills for agents

Audience

IT professionals Hands-on IT leadership

Contact