The HaiPhai Curriculum
The AI-Compressible Biotech
Where your timelines actually live, how to find your organization's highest-leverage AI opportunities, and how to measure compression in outcomes that matter.
NDA in a Week
Why regulatory dossiers take 6 months and don't have to. Parallel architecture, AI-assisted drafting, and human review workflows that collapse timelines without sacrificing quality.
Global Filings in Days
Single-dossier architecture for multi-jurisdiction submissions. Regulatory localization at scale across ICH regions, languages, and format requirements.
Trial Design and Accrual at Scale
Why 40% of trials miss enrollment targets and how to fix it. AI-assisted protocol design, site selection intelligence, patient identification, and accrual monitoring.
From 1,000 Molecules to the Next Candidate
How to use AI to compress the drug discovery cycle. SAR at scale, pharmacophore analysis, generative molecular design, and ADMET-first filtering.
Ops at the Speed of Science
Eliminating the G&A tax. Contract automation, grant intelligence, and building an AI-first back office that stops slowing the science.
Cross-Functional AI Coordination
Why individual AI tools don't compound. Workflow redesign, institutional memory, and the governance structures that turn individual capability into organizational advantage.
Prompting for Biotech Outcomes
Context as a force multiplier. Patterns for scientific documents, regulated environment constraints, chain-of-thought for complex work, and building a team prompt library.
AI Governance for Biotech
What data you can and can't use, how to build a policy that enables instead of blocks, validation and audit trail requirements, and a 90-day governance roadmap.
How Claude Works — Mental Models & Limits
What Claude actually is, how the context window shapes every interaction, when to trust vs. verify outputs, and what data never belongs in any AI tool.
Setting Up Claude Code & Claude Chat
Account setup, plan tiers, installing the Claude Code CLI, IDE integration, writing your first CLAUDE.md, and configuring Claude.ai Projects for your team.
Prompting Claude Well
The anatomy of a high-performance prompt, the four pillars of effective instructions, iterative refinement, building a shared prompt library, and diagnosing failure modes.
Code Coworking with Claude Code
The pair-programming mindset, reading unfamiliar codebases, debugging workflows, refactoring and test generation, safe agentic tasks, and extending Claude Code with hooks and MCP servers.
Chat Workflows for Knowledge Work
Document drafting and editing loops, research synthesis, data analysis with artifacts, meeting prep and async comms, and using Projects for persistent team memory.
Team Conventions & Continuous Improvement
Establishing shared CLAUDE.md conventions, reviewing AI-generated work, onboarding new members, measuring impact, and running retrospectives for your AI practices.
The Yungsten Method — Prototype to Production
How Yungsten turns fragile AI prototypes into systems teams can actually run. Covers our production standard, client intake process, the 90-day engagement arc, and what 'done' means on a Yungsten project.
Building Named Agents with Claude Desktop
End-to-end process for designing, configuring, testing, and handing off named AI agents using Claude Desktop, MCP servers, and system prompts. Four agents per quarter, starting here.
Client Wiki Systems — Obsidian & CLAUDE.md
How to architect, populate, and maintain an executive AI wiki in Obsidian. Writing CLAUDE.md files that teach Claude about a client's organization. Wiki governance and the quiet-tending practice between visits.
Facilitating Client AI Sessions
Running effective executive AI sessions and monthly team seminars. Teaching clients to prompt well, handling resistance, and building the internal fluency that makes Yungsten engagements stick after we leave.
Production Engineering for Agentic Systems
For implementation engineers: production deployment patterns, error handling in agentic workflows, tool failure recovery, and the technical handoff documentation standards that make client systems maintainable at scale.
Yungsten Track Capstone — End-to-End Engagement Simulation
A single 90-minute scenario-based challenge covering the complete Yungsten workflow: intake brief, engagement scoping, agent specification, system prompt, MCP configuration, test plan, operator runbook, wiki entry, CLAUDE.md, and architecture document. Demonstrates mastery of Modules 15–19.