Facilities & Environment

PromptGenix LLC · Contact: dohoon.kim1@icloud.com · promptgenix.org
Overview

Phase I execution is supported by a modern local compute environment and secure, cloud-ready workflows.

PromptGenix Phase I focuses on an evidence-weighted hypothesis intelligence engine that integrates public biomedical datasets and literature. Development is supported by a dedicated local workstation for rapid iteration and deterministic reruns, with an optional secure cloud path (e.g., AWS VPC) for scalability and collaboration testing as needed.

Phase I validation will use public datasets only to avoid IP and data-use constraints, while demonstrating feasibility, reproducibility, and transparent evidence linking suitable for SBIR review.

Local development environment

  • Dedicated workstation: high-memory Mac Studio-class system for rapid prototyping and iterative development
  • Core toolchain: Python/R for evidence feature extraction, inference, and report generation
  • Reproducibility: versioned environments, pinned dependencies, config snapshots, and checksums
  • Artifacts: exportable HTML/PDF reports, evidence objects, and audit-friendly logs
Local-first Deterministic reruns Evidence objects HTML/PDF outputs

Secure & scalable deployment option

  • AWS-ready architecture: supports deployment in a secure VPC for scale testing and controlled collaboration
  • Data separation: Phase I uses public data; sensitive datasets remain external to PromptGenix environments
  • Controlled access: role-based access patterns and audit-friendly storage options (as applicable)
  • Portability: reproducible configs enable consistent execution across local and cloud contexts
Phase I policy alignment: evaluation uses GEO/SRA/FlowRepository and public literature sources to ensure clean SBIR review.
Resources supporting Phase I

Software, public datasets, and operational supports aligned to SBIR deliverables.

Software stack

  • Evidence layer: standardized extraction of effect sizes, uncertainty, reproducibility, and context descriptors
  • Inference: statistical scoring + Bayesian updating for posterior confidence estimates
  • Interpretability: LLM used only for explanation constrained by evidence links (“no evidence → no claim”)

Validation sources

  • Public datasets: GEO/SRA (omics) and FlowRepository (immune profiling)
  • Public literature: PubMed/PMC for priors and traceable citations
  • Evaluation focus: traceability, reproducibility, calibrated uncertainty, and utility

Operational supports

  • Documentation: SBIR portal pages + reviewer-ready evidence-linked reports
  • Pilot readiness: structured onboarding and feedback templates for partner labs
  • Collaboration: mechanisms to support letters of support and Phase II pilot expansion
Facilities fit to the work

Environment supports rapid prototyping, reproducible validation, and SBIR-compliant evaluation.

Need Environment support Phase I output
Fast iteration Local compute enables rapid evidence-layer development, inference tuning, and report refinement Stable end-to-end hypothesis ranking on multiple public datasets with deterministic reruns
Reproducibility Pinned versions, config snapshots, and checksums; exportable evidence objects and logs Rerun-consistent outputs; reviewer-ready artifacts with traceability and uncertainty labeling
Scalability testing Cloud-ready execution path (optional) for heavier workloads and multi-user patterns Feasibility evidence for Phase II secure deployments and multi-site pilots
Bottom line: The current facilities and environment are sufficient to deliver Phase I milestones (evidence ingest, hypothesis prioritization, traceable reporting, and measurable validation on public data), while maintaining a clean path to secure scale-out in Phase II.