Skip to content
View camerontjs-dot's full-sized avatar
  • Toronto
  • 14:31 (UTC -04:00)

Block or report camerontjs-dot

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
camerontjs-dot/README.md

Cameron Sanderson

Pharma QA/QC operator building inspectable AI workflows for regulated and evidence-heavy work.

I have more than eight years of experience across sterile compounding, analytical QC, OOS investigations, controlled records, safety programs, and direct communication with physicians and pharmacists. I now apply the same operating habits to software: preserve the source state, make review points visible, test the failure modes, and leave enough evidence for someone else to challenge the result.

Based in Ontario, Canada. Bilingual English and French.

Start here

If you have one minute, these three repositories explain the direction:

  1. Claim Audit Lab checks whether draft claims are supported by supplied evidence. It uses deterministic rules and produces reviewable Markdown and JSON reports with explicit limits.
  2. Evidence Bundler turns bounded source material into reviewable evidence bundles. Retrieval nominates candidate passages; human review decides what belongs in the final bundle.
  3. Apparatus Contracts defines versioned handoffs, controlled vocabulary, and integrity checks for moving artifacts between tools without hiding the boundary.

Together, they show the pattern I am building toward: source material enters, provenance stays attached, review remains explicit, and final claims can be checked against the evidence the workflow actually supplied.

Read the claim-support workflow case study.

The workflow

flowchart LR
    A["Bounded source packet"] --> B["Evidence Bundler"]
    B --> C["Candidate evidence"]
    C --> D["Human review"]
    D --> E["Reviewed evidence bundle"]
    E --> F["Claim Audit Lab"]
    F --> G["Claim-by-claim audit report"]
    H["Apparatus Contracts"] -. "versioned schemas and integrity checks" .-> B
    H -. "versioned schemas and integrity checks" .-> F
Loading

The public tools are not a truth engine or a regulated quality system. Evidence Bundler nominates and packages evidence. Claim Audit Lab audits support relative to supplied evidence. Apparatus Contracts defines the intended handoff shape. The full contract adapter and upstream research harness remain active work rather than public-release claims.

Other public work

Career Decision Engine is a browser-based decision-support tool with visible scoring, separate rule checks, confidence labels, and validation sweeps. Try the live demo.

MindGraph is a local retrieval engine over a Markdown knowledge workspace. It combines lexical and semantic retrieval, typed document links, bounded graph expansion, and an MCP interface.

Mainframe is the Markdown-first knowledge and project workspace that MindGraph indexes. It separates inbox, ingest, durable knowledge, live state, projects, and archive by lifecycle.

Basic Research Harness compares a raw Python agent loop with a project-configured SDK workflow so the procedure, review step, stop condition, and QA gate remain visible.

What I am looking for

I am most useful where regulated operations meet implementation:

  • regulated software implementation and customer technical services;
  • quality systems, data integrity, and digital quality work;
  • AI evaluation and evidence-handling workflows;
  • forward-deployed or consulting work that requires mapping a real process before automating it.

My software work is self-taught and project-based. I am not presenting myself as a finished senior software engineer or ML researcher. The value I bring is the bridge between regulated operations, reviewable evidence, and working systems.

Contact

Pinned Loading

  1. evidence-bundler evidence-bundler Public

    Traceable evidence-bundle preparation for research corpora, with retrieval nominations, review sidecars, provenance, and adapter-ready outputs. It is a RAG-adjacent evidence-preparation pipeline: r…

    Python 1

  2. claim-audit-lab claim-audit-lab Public

    Supplied-evidence claim audit CLI with traceable Markdown and JSON reports.

    Python

  3. basic-research-harness basic-research-harness Public

    A small agent-engineering harness that turns a research topic into an inspectable JSON bundle with source-to-claim provenance. Built two ways: a raw Python agent loop and a Claude Code SDK rebuild …

    Python

  4. apparatus-contracts apparatus-contracts Public

    Canonical handoff contracts and a hash-verified verifier suite for the research scaffold evaluation apparatus. Regulated-industry data-integrity grammar (ALCOA+, 21 CFR Part 11) applied to AI workf…

    Python

  5. MindGraph MindGraph Public

    Local, graph-augmented retrieval engine for personal Markdown knowledge bases. One SQLite file, no service. Fuses BM25 + sqlite-vec with RRF and walks a typed [[link]] graph.

    Python

  6. MainFrame MainFrame Public

    Markdown-first knowledge workspace organized by information lifecycle. Inbox → ingest → knowledge → live → projects → archive, each with its own update rules. Deterministic routing, source-backed e…

    Python