Manifold Grace
Manifold Grace · Grace

Software Engineering

A technical portfolio for code, infrastructure and engineering reasoning. The emphasis is on explaining what was built, why particular choices were made, what failed and what was learned — not simply displaying repositories.

TECHNICAL PORTFOLIO

Builds with explanations

Projects can be presented as compact engineering narratives: problem, architecture, implementation, security considerations, testing, failure modes and next iteration. Code blocks are useful where they illuminate a decision, but the page should remain intelligible to someone who is not reading an entire repository.

This format also supports ethical analysis. An algorithm or infrastructure choice can be examined technically and then connected to questions of observability, data use, access, reliability, power or governance.

Published development writing

My DEV Community writing records participation in software-development discussions and the continuing relationship between practical engineering, systems thinking and responsible technology.

Read: Excited to Join the Developer Community (DEV Community)

Current engineering themes

Current development includes cloud and DevOps architecture, networking, containers, Terraform, GitHub Actions, observability and AI infrastructure, alongside the audiovisual systems knowledge accumulated through professional practice.

For doctoral or advanced research, technical write-ups can become evidence for analysing how engineering decisions shape ethical and institutional outcomes rather than treating ethics as an afterthought applied once a system is already built.

AI METHODS · TAXONOMY

1,000 AI method and application combinations

An interactive list-only reel that enumerates 40 AI families × 25 study lenses = 1,000 indexed combinations. These are labelled combinations for exploration, not a scientific claim that 1,000 fundamentally distinct AI types exist.

No opacity fades. Reel motion pauses when the page is hidden or reduced motion is requested; use the controls to navigate manually.

Open the AI Research Laboratory to explore runnable examples and architectural logic →

INTERACTIVE · AI MODEL & METHOD REEL

1,000 AI methods, architectures & applications

Browse 200 named techniques and 800 domain-specific application patterns. These are not 1,000 mutually exclusive scientific definitions of AI; the reel distinguishes actual techniques from illustrative applications. It lists terms without explanations.

METHOD001 / 1000
Expert systems
Symbolic and logic
Open searchable alphabetical / category directory

    Playback is user-controlled; automatic motion is disabled if reduced motion is preferred. No opacity fading or image crossfades.

    PLATFORM STRUCTURE · ANNOTATED

    Anatomy of an AI platform

    Follow the chain from researcher to result. Each level is independently testable in the interactive AI Research Lab.

    1. 01 · Research interface

      Browser UI, experiment parameters, source and code explorers, result visualisation. Never place confidential API tokens in frontend code.

    2. 02 · Identity & authorisation

      Session, MFA where available, organisation roles, per-project permissions, and consent / licensing gates.

    3. 03 · API gateway & orchestration

      Request schema validation, rate limits, routing, prompt/tool policies, cost caps and tracing IDs.

    4. 04 · Retrieval & data layer

      Document ingestion, metadata catalogue, vector search, embeddings and provenance. Partition confidential and public datasets.

    5. 05 · Model execution

      Inference endpoints, model versioning, evaluation sets, hardware accelerators and sandboxed compute.

    6. 06 · Evaluation & monitoring

      Accuracy, fairness, robustness, quality tests, performance, logs, alerts, incident records and audit trails.

    7. 07 · Reproducible output

      Versioned datasets, reproducible notebooks, model cards, source citations and human review before publication.

    How the security boundaries relate

    User → authenticated frontend → authorised API → isolated compute worker → restricted storage → reviewed result. A visible UI restriction is never a substitute for server-side authorisation.

    OWASP security principles · NIST AI risk management