Disciplined orchestrationbeats bigger models.

The first open end-to-end system for direct ANE inference and training — bypassing CoreML to program Apple's Neural Engine directly.

An empirical microbenchmark study that reverse-engineers Apple's Metal 4.1 tensor compute path — the Metal Performance Primitives matmul2d operation — on the M4 Max GPU, recovering eleven hidden or contradicted hardware specifications Apple's documentation omits.

A quantitative model for post-hoc cooperative LLM training — predicts fusion gain before any compute is spent. Contributors fine-tune checkpoint copies independently; KALAVAI fuses them via lightweight MoE routing in 500 steps.

A formal characterization of how multi-agent AI systems collapse into self-reinforcing consensus — and the four metrics that predict it.

Measurement framework for multi-model orchestration using self-consistency variance routing. Evaluated on 1,510 tasks across 4 benchmarks with 7,550+ auditable runs.

A host-agnostic refinement protocol for multi-agent AI research — five specialist agent roles, strict file-first workflow, anti-coherence-trap mechanisms, and fully auditable experiment records. Published on npm and actively used in production.

A factorial decomposition of an evolutionary mixture-of-LoRA system — isolating whether the win comes from the router rewrite, the per-domain leave-one-out evaluation scope, or the lifecycle (death + α-blend inheritance + SVD mutation + slot reallocation).

An LLM architecture for uploadable machine learning with architectural composability guarantees. Any lab trains a specialist above the published frozen Foundation Spine; the SANGAMAM Composability Interface (SCI) is the formal contract that guarantees specialists compose correctly without central coordination. Builds on the empirical fusion-gain proof from KALAVAI — but makes composability *designed*, not accidental.

A living world model and knowledge operating system for documented Tamil heritage sites. Phase 2 extension of THIRUVAASAL — generalizing single-site probabilistic reconstruction into a multi-site, evidence-aware, versioned, queryable infrastructure. Exposes evidence submissions, reconstruction posteriors, provenance, temporal slices, and confidence fields as API contracts before any generation happens.

A pre-registered, cheap-baseline screening rule that decides — before any expensive evolutionary or population-based outer loop is built — whether that outer loop can possibly help. A single Phase-0 gate computes a recovery ratio R; when R ≥ 90%, the outer loop is predicted to add nothing and is skipped.