Skip to content
View mgillr's full-sized avatar
9 followers · 62 following

Block or report mgillr

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
mgillr/README.md

Ryan Gillespie

PMO & programme management — complex integrated systems and capital delivery. Engineering background in rail signalling and safety-critical commissioning; building open-source tools for deterministic distributed AI on the side.

Switzerland · data@optitransfer.ch

I started hands-on — electrical, control and signalling systems, test and commissioning on safety-critical railways (ERTMS/ETCS, ATP, TCMS) — then moved into PMO stand-up and programme controls. Recent programmes: Crossrail (Elizabeth line), Lilium, and Roche's first CGM. Somewhere along the way the engineering pulled me back in.

What I work on in my own time

crdt-merge — a merge library where every strategy converges: dataframes, JSON and model weights, merged without a coordinator. Python, TypeScript, Rust and Java editions. Paper: arXiv:2605.19373.

acfa-rs — Byzantine-robust federated aggregation with receipts anyone can re-execute. Open, Apache-2.0. Paper: arXiv:2607.10305.

Model merging — a long-running experiment in folding fine-tunes and whole architecture families into a single model, gated so it only keeps changes that hold up on measurement. Collectives published under Optitransfer on Hugging Face.

Rail — a research project on onboard train integrity for pure ETCS Level 3: what carries the safety argument once the track circuits go. Machine-checked where it can be, honestly approximate where it can't (yet). Patents GB 2607132.4 / GB 2608127.3 filed.

Always learning in public — issues and pull requests welcome on any of it.

Pinned Loading

  1. crdt-merge crdt-merge Public

    Conflict-free merge for DataFrames, JSON, ML models & distributed agents — powered by CRDTs. The first merge library where every operation is mathematically guaranteed to converge.

    Python 7

  2. acfa-rs acfa-rs Public

    Train one machine-learning model across many parties who don't trust each other. A dishonest minority can't skew the result, and every party computes byte-for-byte identical output on any CPU or ar…

    Rust

  3. memora-swarm memora-swarm Public

    Bit-exact Byzantine-fault-tolerant state layer for federated learning and edge/defence swarms — client SDK, framework adapters, docs & validation battery for the Memora consensus engine.

    Python