Operational doctrine for practical AI systems design.
AI companion for your workspace, with per‑project memory, coding agents, and a native desktop chat UI
[EuroSys 2026] TokenFlow: Buffer-aware preemptive scheduling and proactive KV-cache management for responsive LLM streaming.
The AI Contract Runtime
Perpenda — Android learning path for product professionals learning LLM systems through trade-offs, rubrics, and calibrated AI grading.
A muti-agent Dnd DM that directs you to dive into a fantasy world!
PyTorch DDP and multi-GPU LLM training — from a minimal distributed example to NanoGPT speedruns, Muon, H100 profiling, and modded-nanogpt. FBA LAB https://bubblnet.com
Structure-first pipeline that transforms GitHub repositories into question-indexed, graph-grounded semantic assets for code search, Graph-RAG, and LLM-based reasoning.
AI-Powered Ad Campaign Orchestration Platform
Documentation and design artifacts for Channel AI, a multi-agent conversational analytics system.
Open-source harness distillation: frontier teachers improve prompts, tools, validators, skills, and runtime policies around weaker models.
PromptGuard is a pragmatic, opinionated framework for establishing continuous integration for LLM behavior. It operates on a simple, verifiable principle: run the same prompts across multiple model configurations, compare outputs against defined expectations, and flag semantic regressions.
nanhudev / Xuanjun Yu — AI-native builder working on agent systems, Brain x Harness infrastructure, local-first AI and creative tooling.
A unified, production-grade LLM training and alignment framework. Features modular implementations for downstream tasks BERT, DPO, and GRPO to streamline model alignment and optimize memory overhead.
Automatically extracts, validates, and structures invoice data from images and PDFs using OCR + a local LLM—eliminating manual data entry.
IntelliMesh AI is an **agentic decision system** designed for enterprise AI architectures—integrating **decision intelligence**, **multi-agent orchestration**, and **document intelligence pipelines** into a unified execution layer.
Research-grade neuro-symbolic RAG framework where retrieval is a policy, not a vector search, built for evaluation, ablation, and reliability analysis.
A hands-on technical book and Python lab for building and evaluating agentic systems.
Prototype for execution-plan-driven data standardization, schema profiling, validation, and preview generation.
To associate your repository with the llm-systems topic, visit your repo's landing page and select "manage topics."