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.

Reproducible Python platform for methane-oriented sensor streams, statistical change detection, drift analysis, SQLite persistence, FastAPI, Docker, CI, and benchmark evaluation.

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）

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

Agent-guided coding interview prep for any role: training workflows, original exercises, Python test helpers, and VS Code setup for industrial, algorithmic, and ML interviews.

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

Public portal and research gateway for an edge-first AI research system spanning agent research, evidence engineering, research infrastructure, digital archaeology, and deterministic knowledge lifecycles.

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.

A closed one-person research lab on crypto perpetual-futures microstructure, June to September 2026. No edge found.

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.

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

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

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