WORK & EXAMPLES
Three kinds of thing, kept apart on purpose.Three kinds of thing, kept apart on purpose. Nothing here is a customer case study, and every record carries the verification it still needs.
Systems we built and can show end to end. Capability evidence, not a customer outcome.
Repository material describes e-commerce data collection. A second repository is substantially upstream framework example documentation.
STILL TO VERIFY
Code provenance, implemented custom functionality, licensing and attribution, current runnable state, and sanitised demo data. Repository presence alone does not establish originality or production maturity.
RECORD: DRAFT - NOT CLEARED FOR PUBLICATION
Explicitly a personal research project: data collection, evaluation and operations. The portfolio reports the research did not establish a robust trading edge.
STILL TO VERIFY
Which engineering artifacts can be shown and reproduced. Frame as engineering and evaluation experience; do not imply profitable trading or client outcomes.
RECORD: DRAFT - NOT CLEARED FOR PUBLICATION
Read-only, source-aware analytics assistant over operational entities, with report outputs, citation and freshness events, restricted database roles, long-query handling, and regression tests.
STILL TO VERIFY
Implementation and capability evidence, not a customer transformation or measured business result. Verify ownership, branch and deployment status, approved screenshots, and which features are safe to show. Do not expose real tenant, vendor, SKU, employee or financial data.
RECORD: DRAFT - NOT CLEARED FOR PUBLICATION
Document ingestion, semantic chunking and embeddings, retrieval with source citations, no-reference responses, structured context-to-risk suggestions, and configurable impact and frequency assumptions.
STILL TO VERIFY
Describe AI output as a reviewable suggestion, not a risk conclusion or compliance or legal advice. Verify production status and permission boundaries before calling it a deployed system.
RECORD: DRAFT - NOT CLEARED FOR PUBLICATION
Backend history supports cited corpus chat and a document-analysis direction with findings and dates. The associated mockup explicitly states its screens use fictional data and have no backend, database or AI.
STILL TO VERIFY
Separate implemented code from concept mockups. A document-review scenario may be an owned demonstration, but no automated legal conclusion, customer result, or production claim is supported by the mockup alone.
RECORD: DRAFT - NOT CLEARED FOR PUBLICATION
Work done personally, before Subspace Works, described without naming a protected employer or client.
Natural-language analytics and a supporting data platform, described in the personal portfolio; a dated 2024 launch article describes the product as an alpha.
STILL TO VERIFY
Reconcile dates and maturity; verify ownership, individual contribution, permission, and any production or user claims. Do not reuse the portfolio's speed or scale figures without their measurement records.
RECORD: DRAFT - NOT CLEARED FOR PUBLICATION
Portfolio records contract work on documentation tooling, code generation, and developer education.
STILL TO VERIFY
Accurate role wording, public corroboration or permission, and any claimed numerical outcomes. Describe the principal's previous contribution, not a Subspace Works client engagement.
RECORD: DRAFT - NOT CLEARED FOR PUBLICATION
Portfolio records mobile and web engineering and API work.
STILL TO VERIFY
Same role, permission, attribution and metric checks. Relevant supporting experience; weaker fit for the two initial offers.
RECORD: DRAFT - NOT CLEARED FOR PUBLICATION
Documented PDF-to-product-data conversion, direct device upload, printable lists, and compatibility testing.
STILL TO VERIFY
Store-owner permission, current implementation checks, sanitised inputs, and actual operating outcomes.
RECORD: DRAFT - NOT CLEARED FOR PUBLICATION
Synthetic demonstrations authored to make an offer concrete. No client engagement and no measured business outcome.
Repeated exports, reconciliations and reporting steps consume expensive human effort.
WHAT WE CAN BUILD TODAY
ERP has the strongest directly relevant feature set; some edge cases and source-freshness issues remain open.
SYNTHETIC - NO CLIENT ENGAGEMENT OR MEASURED OUTCOME
A feature or prototype exists but quality is inconsistent and release criteria are undefined.
WHAT WE CAN BUILD TODAY
Model seam, eval/cost harness, permission controls and reliability work are useful evidence even where issues remain.
SYNTHETIC - NO CLIENT ENGAGEMENT OR MEASURED OUTCOME
Answers exist across a corpus nobody has read end to end, so people ask colleagues instead.
WHAT WE CAN BUILD TODAY
Ingestion, semantic retrieval, citations and no-reference behaviour exist; broader enterprise assistant is not proven.
SYNTHETIC - NO CLIENT ENGAGEMENT OR MEASURED OUTCOME
Document-heavy teams repeat a review workflow where the exceptions are what matter.
WHAT WE CAN BUILD TODAY
Backend history points toward cited findings and dates; the mockup is explicitly fictional and non-AI.
SYNTHETIC - NO CLIENT ENGAGEMENT OR MEASURED OUTCOME