Writing
Notes on LLM agents, clinical trial programming automation, and CDISC data standards — extended companions to my published research.
State of the Field
Quarterly surveys of the evidence — what's proven, what's hype, with full references.
2026-08-30
A tiered survey of LLM evidence in clinical trial statistical programming: benchmarked results, promising single-team studies, vendor hype, and open gaps for 2026–2027.
#llm#clinical-trials#statistical-programming#survey#benchmarks#gxp
Field Guides & Hot Topics
Practical explainers on the tools and debates shaping clinical programming today.
2026-08-30
How to tell demo-ware from deployable agentic AI in clinical trials: determinism, audit trails, 21 CFR Part 11, constrained architectures, and vendor questions.
#llm-agents#clinical-trials#gxp#audit-trail#agent-architecture
2026-08-30
Where ChatGPT, Claude, and Copilot genuinely save time in SAS and R clinical programming, where they fail, and how to use them inside auditable GxP workflows.
#ai-coding-assistants#sas-programming#gxp#clinical-programming#llm-tools
2026-08-30
CDISC CORE is a free, open-source rule engine for SDTM and ADaM validation. How its YAML rules work, how to run it, and where rule-based checking stops.
2026-08-30
Why pharma is moving from SAS to R, what breaks in practice (procedures, XPT files, QC), and how to migrate a validated macro library without a rewrite.
#sas-to-r#pharmaverse#admiral#clinical-programming#xpt#validation
Notes
Short, frequent observations — tools, papers, and field notes. Longer arguments live in the sections above.
2026-08-30
Setting temperature to zero feels like determinism. It isn't — and in GxP work the difference will find you during an audit, not during development.
2026-08-30
69.7% of oncology CORE rules port cleanly to a graph constraint language. The 30% that don't are exactly where the dangerous errors live.
Research Deep-Dives
Extended companions to the published papers — methods, results, and honest limitations.
2026-08-27
Rule-based SDTM validators structurally miss cross-domain contradictions. SHACL-SPARQL graph constraints plus a deterministic agent layer catch all 20 archetypes.
2026-08-25
Independent QC re-programming costs 30–50% of clinical programming effort. An AI framework matched 97.1–100% of variables while keeping independence intact.
2026-08-22
Five LLM generation methods benchmarked across 1,999 bootstrap experiments on ICH E3-conformant TLF templates: hybrid RAG with reranking beats direct prompting.
#rag#llm-benchmarking#tlf-templates#clinical-trials#r-code-generation
2026-08-20
Free-form agent loops break down in GxP clinical programming. Structuring the work as a typed process DAG makes LLM agents reliable, replayable, and auditable.
2026-08-19
A GRADE-rated review of 2020–2025 automation evidence in statistical programming: real gains, mostly Low to Very Low quality evidence.
#statistical-programming#automation#evidence-quality#clinical-trials
2026-08-16
A bridge map, typed IR, and orchestrator wrap a legacy SAS TFL library unchanged — AI-ready JSON on day one, 80%+ cell-level parity, optional 92% code cut.
#sas#legacy-modernization#clinical-trials#llm-integration#statistical-programming
2026-08-13
ClinAgent splits clinical programming capability into five layers, keeping MCP tools stateless and packaging domain expertise as testable skills.
#llm-agents#mcp#clinical-trials#statistical-programming#agent-architecture
2026-08-10
Schema-only synthetic ADaM generation plateaus at 0.45 overall quality; enriching schemas from protocol/SAP/CRF knowledge graphs plus templates reaches 0.70.
#synthetic-data#adam#knowledge-graphs#llm#statistical-programming
2026-08-07
Base Llama 3.1 8B scores 0.36 on admiral code generation. LoRA fine-tuning plus knowledge-graph validation gets it to 0.82 — without sending data to an API.