Services

Apache MXNet moved to the Apache Attic in 2023. There are no new releases, no security patches, and no maintainers on call. If your models, training pipelines, or inference services still depend on it, the framework is not going to fail loudly — it is going to rot quietly under you: unpatchable CVEs in the dependency tree, CUDA and driver versions that no longer line up, base images you can't rebuild.

NeuraBloq exists for exactly this situation. We keep MXNet systems running safely while we move them somewhere maintained. The work falls into four engagements, usually in this order:

MXNet estate assessment

Before anything moves, know what you have. We inventory every model, training pipeline, and serving path that touches MXNet, map dependency and CVE exposure, and document the hardware and driver constraints each one lives under. The output is a per-model recommendation: port it, contain it, or retire it — with the reasoning written down.

Model porting

Re-implementing and re-validating models in PyTorch, or ONNX-first where that is the better fit. Weight conversion where it holds, numerical-parity testing to prove it holds, and honest retraining estimates where conversion won't be faithful. A port isn't done when it compiles; it's done when the parity harness says the new model is the same model.

Pipeline & serving migration

Models are only half the estate. Gluon training code, SageMaker MXNet containers, and MXNet-based inference services all need maintained replacements. We build the new pipeline alongside the old one and run them in parallel — same inputs, compared outputs — before anything is cut over.

Containment support

Some systems can't move yet: budget, staffing, or a model nobody fully understands anymore. For those we build a containment posture — pinned environments, vulnerability monitoring, network isolation — and a dated exit plan, so "not yet" never becomes "never."

How engagements run

All work is hourly time and materials, staffed by senior US engineers, as project engagements or staff augmentation. Scope, staffing, and timeline are agreed through an initial conversation — get in touch and tell us what your MXNet estate looks like.