research-engineering

Research prototype for trace-based observability and failure analysis in retrieval-augmented generation.

Autonomous AI research team: write a plan.md, spawn agents, train, verify, and deliver a clean ML repo.

C++/Python microstructure research engine for event-driven limit order book prediction and reproducible quantitative experiments.

A gated memory layer for trustworthy AI-assisted research workflows.

Real-time TUI for monitoring cloud GPU training instances

Automates hermetic environments (macOS/HPC) to eliminate drift. Provisions offline RAG (Gemma 2), compiles LaTeX manuscripts, and indexes local knowledge. Unifies infrastructure, writing, and inference into a single, audit-ready artifact.

Skill中间件架构：AI时代研究能力的工程化封装（以固定收益研究为例）

Research-to-production AI model release control plane connecting post-training experiments, baseline/candidate evaluation, regression analysis, safety/correctness checks, CI release gates, production telemetry, and evidence-linked SHIP · INVESTIGATE · HOLD decisions.

A Codex plugin that reduces over-design, hidden failure, and evidence inflation in research engineering.

Research-engineering portfolio: LLM evaluation, RAG, RLVR/post-training, machine learning, Rust systems, data engineering, and reproducible computation.

Gated research harness: commit and submit gates, provenance, cluster submission — the scaffolding with the project removed

Selected LLM, NLP, retrieval and vision-language research-engineering projects.

Factor-aware, physics-guided road-surface intelligence | 因子感知与物理引导的路面状态识别、摩擦可供性估计与可复现实验

From-scratch diffusion models research repo. Implements DDPM, DDIM, ODE samplers, UNet with attention, LoRA, EMA, CFG, and original adaptive sampling research. Includes paper summaries, math derivations, benchmarks, and experiments.

Reference model for managing state, intent, evidence, provenance, and execution in long-running agent-assisted research engineering.

Event-driven quantitative research evidence engine with real-data risk studies, leakage-safe execution, corrected inference, and reproducible reports.

Evidence-first A-share bottom-position and intraday-T research: leakage controls, risk filters, transaction costs, rolling OOS validation, and audit governance.

面向竞赛与实验论文的可审计、可复现工程化 Skill：实验评测、不可变基线、证据链与存量项目接管

Catch evaluation drift before it ships.

AI systems, quantitative research, and governed agent infrastructure.