A Zero-Dependency, Browser-Native Compute Pool for Heterogeneous and Untrusted Devices

This technical report presents a pooled-compute system in which unmodified web browsers and headless Node.js processes pull small, self-contained tasks from a coordinator that never executes a task itself. The whole system, including the coordinator, the device runtime and every numerical kernel, is written without third-party dependencies. It runs data processing, tabular and text machine learning, small sequence models, optimisation solvers, agent flows, real large-language-model inference and LoRA/QLoRA fine-tuning under one scheduler and one credit ledger. The report describes the design and a preliminary evaluation: a pull-based task protocol with retries, backups and cross-checks; a data-parallel local-SGD engine with shard-only data, quorum pacing and a peer check of every update; a layer-pipelined LLM engine with a learned latency model, a fused LM head, optional WebRTC links and verification probes; an acyclic, signed credit ledger; and coordinator protection. Measurements come from a single PC with Qwen2.5-0.5B. They are small-scale, and the report states clearly what they do not show. Most ingredients are adapted from published work; the contribution is their integration into one browser-native system.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-13
DOI
https://doi.org/10.5281/zenodo.24184358
Primary Topic
Distributed and Parallel Computing Systems
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A Zero-Dependency, Browser-Native Compute Pool for Heterogeneous and Untrusted Devices

Ashlin Gabriel Rajan
1 links
Zenodo (CERN European Organization for Nuclear Research)
Distributed and Parallel Computing Systems
article

A Zero-Dependency, Browser-Native Compute Pool for Heterogeneous and Untrusted Devices

Ashlin Gabriel Rajan
article en

Abstract

This technical report presents a pooled-compute system in which unmodified web browsers and headless Node.js processes pull small, self-contained tasks from a coordinator that never executes a task itself. The whole system, including the coordinator, the device runtime and every numerical kernel, is written without third-party dependencies. It runs data processing, tabular and text machine learning, small sequence models, optimisation solvers, agent flows, real large-language-model inference and LoRA/QLoRA fine-tuning under one scheduler and one credit ledger. The report describes the design and a preliminary evaluation: a pull-based task protocol with retries, backups and cross-checks; a data-parallel local-SGD engine with shard-only data, quorum pacing and a peer check of every update; a layer-pipelined LLM engine with a learned latency model, a fused LM head, optional WebRTC links and verification probes; an acyclic, signed credit ledger; and coordinator protection. Measurements come from a single PC with Qwen2.5-0.5B. They are small-scale, and the report states clearly what they do not show. Most ingredients are adapted from published work; the contribution is their integration into one browser-native system.

Zenodo (CERN European Organization for Nuclear Research)
Distributed and Parallel Computing Systems
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon