Skip to content
View sparkainlp-x's full-sized avatar
0 followers · 1 following

Achievements

Achievement: YOLO

Achievements

Achievement: YOLO

Block or report sparkainlp-x

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.

Content in all repositories owned by your account will be closed.
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
sparkainlp-x/README.md

Spark AI NLP banner: glowing brain-circuit logo with the words SPARK AI NLP over a dark lab scene

Spark AI NLP

Disability-led research software from Fredericton, NB, Canada.
RAG guardrails, residual-latch anomaly detection, C++/HLS research prototypes, and evidence-tagged artifacts.

sparkainlpx.xyz · LinkedIn · Founder: Jean-François Brisson · Français (langue maternelle) · English (fluent)

Research prototyping C++ and Python HLS and FPGA: TARGET (design goal, synthesis not yet run) Evidence-tagged research

About

Spark AI NLP is a disability-led research-software company based in Fredericton, New Brunswick, Canada. The public repositories below are research prototypes: a source-grounded RAG guardrail, deterministic residual-latch anomaly detection, and C++/HLS exploration. They are not hardware products, field products, or medical products, and they do not report results on quantum hardware.

Featured work

Repository What it is Evidence status
spark-rag-guardrail Source-grounded RAG (ChromaDB + Ollama) that refuses instead of guessing when retrieval relevance is too low 10 tests pass in CI on SYNTHETIC fixtures; answer quality UNRUN
oes512q-latch OES-32/OES-512 classical residual latch with a self-calibrated threshold; 47-series NAB benchmark (mixed results, reported as measured) Unit tests SYNTHETIC; NAB numbers REPORTED; not a quantum code
oes32-residual Normative OES-32 residual (ADR-001): max-absolute residual over two 32-vectors, strict > tolerance rule, contract tests. Release v0.1.1 12 contract tests pass in CI (Python 3.11–3.13); SYNTHETIC inputs
oes32-membrane-shield Ed25519 capability-gated 32-slot state machine with dual-approval calibration and the ADR-001 residual latch 25 tests incl. fuzz pass in CI; SYNTHETIC; independent security review of v2 UNRUN
qldpc_decoder_cpp C++/HLS qLDPC decoder scaffold: CMake + Catch2 + GF(2) micro-benchmark, Vitis/Vivado/PetaLinux scripts Software build and tests pass in CI; BP kernel is a placeholder; hardware results UNRUN; ~70 ns GF(2) micro-benchmark REPORTED (host unspecified), not decoding latency
Other public repositories
Repository What it is Evidence status
oes32_engine Profile A sidecar: standard-library Python telemetry-triage harness (residual latch, EVEN/ODD symmetry, FOLD8) with a technical specification 7 unit tests pass in CI (Python 3.11/3.12); SYNTHETIC inputs; not certified control software
oes32-hls Profile A sidecar: C++ HLS prototype of the same checks; AMD ZCU111 as a design target g++ testbench passes in CI (8/8 checks); FPGA synthesis and latency UNRUN / TARGET
oes512-residual OES-512 = 16 × OES-32 residual blocks: weighted-latch reference and browser avatar Seed-42 benchmark SYNTHETIC; operational use UNRUN
phmt4-montecarlo Heuristic classical numpy simulation (density-matrix "membranes", mirror, Observer/Calibrator), v1 vs v2 ablation 18 tests pass in CI; all numbers SYNTHETIC; not a consciousness or quantum-hardware claim
quantum-error-correction-demo Classical Python toy simulation (no qubits, no QEC code) All outputs SYNTHETIC; seeded smoke run in CI

Every public research repository listed above is archived on Zenodo with a DOI (see each README) and grouped in the Spark AI NLP Zenodo community.

How I work

  • Contracts first. Each definition lives in one normative place with executable tests. For OES-32 that is oes32-residual@b77b612 (ADR-001); other implementations are labelled Profile A sidecars and document how they differ.

  • Fail-closed. Invalid input is rejected rather than guessed, and so are claims: if something has not been run or measured, the README says so.

  • Evidence tags on every number.

    Tag Meaning
    SYNTHETIC Produced by our own simulation or tests on generated inputs
    REPORTED Measured by us; not independently reproduced. The README gives the host and conditions, or states plainly that they are unspecified
    TARGET A design goal; not yet achieved or measured
    UNRUN Tooling or scripts exist; the run has not been performed

  • Reproducible by default. Each README's Quickstart was run from a fresh clone or is marked UNRUN, and repositories with code run their tests in CI on every change.

  • Accessibility first. Ask for alternative formats by opening an issue in any repository.

Contribution guidelines, the code of conduct, and the security policy are shared across all repositories: sparkainlp-x/.github.

Contact

Open to research groups, FPGA/decoder engineers, and accessibility-focused partners.

Popular repositories Loading

  1. quantum-error-correction-demo quantum-error-correction-demo Public

    Classical Python toy simulation of OES-32-style metrics (no qubits, no QEC code). All outputs SYNTHETIC.

    Python 1

  2. oes32-residual oes32-residual Public

    Normative OES-32 residual reference (ADR-001): deterministic max-abs residual over two 32-vectors, strict > tolerance failure rule, contract tests. SYNTHETIC inputs only.

    Python 1

  3. qldpc_decoder_cpp qldpc_decoder_cpp Public

    C++/HLS qLDPC decoder scaffold: CMake + Catch2 + GF(2) micro-benchmark, Vitis HLS/Vivado/PetaLinux scripts for ZCU111. BP kernel and HIL verification are placeholders; hardware results UNRUN.

    C++ 1

  4. spark-rag-guardrail spark-rag-guardrail Public

    Source-grounded RAG (ChromaDB + Ollama) with a retrieval-relevance guardrail that refuses instead of guessing

    Python 1

  5. oes32-membrane-shield oes32-membrane-shield Public

    OES-32 membrane shield: Ed25519 capability-gated 32-slot state machine with dual-approval calibration and ADR-001 residual latch. Research software; SYNTHETIC tests; not certified.

    Python

  6. oes32-hls oes32-hls Public

    Profile A sidecar to oes32-residual: C++ HLS prototype of OES-32 checks for AMD ZCU111. g++ testbench only; FPGA synthesis UNRUN.

    C++