Independent Quantitative Research 獨立量化研究
Daily US equities market insights — powered by machine learning & Agentic AI, published every trading day. 每日美股市場深度洞察報告——機器學習與智能體AI驅動，每個交易日準時發佈。
K2J AI Quantitative Lab is a daily, full-market multi-factor research system for US equities — covering regime classification, breadth analysis, statistical attribution, and dynamic risk management — built end-to-end with production-grade agentic AI. K2J AI量化實驗室是一套覆蓋美股全市場的每日多因子研究系統——涵蓋大盤體制識別、市場廣度分析、統計因子歸因與動態風險管理——基於生產級 Agentic AI 端到端構建。
AI Engineering Stack: Built and generated end-to-end with LangGraph multi-agent orchestration, Agentic RAG, Pinecone vector search, and Deterministic Guardrails. 底層技術棧：基於 LangGraph 多智能體編排、Agentic RAG、Pinecone 向量檢索與確定性護欄（Deterministic Guardrails），全流程端到端自主構建與生成。
How each issue is built報告的構成與方法論
Every issue follows the same five-layer pipeline, constructed daily from post-close US equities market data. 每一份報告都遵循相同的五層流程，基於每個交易日美股收盤後的全市場數據深度構建。
Market regime大盤體制判斷
Classifies risk-on, risk-off, or transitional conditions before anything else is read.在解讀其他信號之前，先判斷大盤所處的體制——風險偏好、風險規避，或過渡階段。
Market breadth全市場廣度
Measures how many stocks are actually participating, not just the index-level headline.衡量真正參與行情的個股廣度，而不只是指數層面的表面數字。
5-D attribution五維穩健歸因
A five-dimensional, noise-robust statistical attribution explains what is actually driving returns.五維穩健統計歸因，剔除噪音，解釋真正驅動收益的因素。
Theme clusters上漲/下跌主題簇
Clusters the day's movers into coherent themes, not a loose list of tickers.將當日漲跌個股歸納爲連貫的主題簇，而非零散的股票列表。
Portfolio & risk組合與風控
Translates the analysis into a portfolio stance and concrete risk-management guidance.將以上分析轉化爲組合策略立場與具體的風控建議。
8 Core Data Pillars Across 100+ Metrics for 1,500+ Liquid Equities 覆蓋 1,500+ 只高流動性核心標的的 8 大維度 100+ 項底層全量數據網絡
A reliable quantitative report cannot rely on shallow percentage changes. To ensure statistical rigor and trade execution, K2J filters the market for top liquidity and scans across 8 core categories: Asset Metadata & Consensus, Price/Volume Microstructure, Technical Oscillators, Financial Statements (TTM & FQ), Advanced Cash Flow Multiples, Growth Rates, 4 Classic Forensic Models (Piotroski / Altman Z / Sloan Ratio % / Zmijewski), and Shareholder Yield. 一份具備真實參考價值的量化報告，絕不可能只看一兩個漲跌幅數字。K2J 在每個交易日收盤後，對全市場精選出的 1,500+ 只高流動性核心標的採集並計算 8 大核心維度：基礎與分析師共識、量價微觀結構、全週期技術指標、財務三大表深度數據、先進現金流估值倍數、複合增速、法務排雷四大經典模型（Piotroski F / Altman Z / Sloan Ratio % / Zmijewski）以及股東回報。
Strict Top-Down "Macro → Sector → Equities" Analytical Flow 嚴密的「宏觀 → 板塊 → 個股」三層推導與逆勢背離預警
Never analyze a stock in isolation. Our daily reports guide you through macro wind direction first, smart money sector migration second, and individual factor attribution third — complete with automated Sector Divergence black-swan warnings. “不謀全局者，不足謀一域”。K2J 研報引導投資者先看宏觀體制明確順逆風與倉位預算，再看產業主題簇追蹤萬億資金跨行業遷徙，最後深入 20 支標的 5 維歸因，並配備獨創的「板塊逆勢背離」量化破案機制。
Macro Regime大盤體制與風險預算
Evaluate benchmark ETFs and 1,500+ liquid stock breadth to determine market climate.判定大盤體制與四大指數走勢，評估 1,500+ 核心標的多空廣度，定總倉位水位。
Thematic Migration產業主題簇與資金輪動
Track institutional smart money rotation across emerging industry clusters.無監督聚類提取領漲領跌主題簇，鎖定機構資金合力抱團賽道。
20 Focus Equities20 股深度歸因與破案
5-D attribution, Causal News Intelligence, and Divergence alerts.5 維共振鎖定 20 支代表，深度推理異動因果並預警逆勢背離閃崩。
Real-Time Sentiment + 13F / SEC Filing Surveillance — Track Smart Money Before It Moves Markets 實時輿情監控 + 13F/SEC申報解析——在聰明錢異動前率先捕捉機構動向
Beyond pure quantitative factors, K2J deploys an NLP-powered Causal News Intelligence engine that continuously monitors and ingests 70+ authoritative financial wire services, Tier-1 business media, and regulatory filing feeds — including real-time parsing of SEC EDGAR submissions: 13F institutional holdings disclosures, 13D/13G beneficial ownership filings, Form 4 insider transactions, and 8-K material event notices. For each liquid equity, the engine computes multi-dimensional event–ticker relevance scoring, bullish/bearish sentiment polarity decomposition, and event-driven tail-risk signal extraction. 13F quarterly snapshots and interim amendments are cross-referenced against price–volume microstructure to surface divergences between disclosed institutional positioning and actual market behavior — transforming regulatory filings and unstructured news flow into a structured alpha overlay enriched with smart-money flow awareness. 在量化因子體系之上，K2J 獨創性地部署了一套 NLP 驅動的因果新聞情報引擎：對 70+ 家權威財經通訊社、主流商業媒體及監管公告信源進行 7×24 小時全天候實時抓取與深度解析。與此同時，系統持續接入並解析 SEC EDGAR 全量申報數據，涵蓋：13F 機構持倉季報與中期修正、13D/13G 大額持股變動申報、Form 4 內部人買賣交易以及 8-K 重大事件公告。針對每一隻個股標的，引擎執行多維度「事件–標的因果關聯度評分」「多空情緒極性解構」「事件驅動尾部風險信號提取」，並將 13F 機構持倉披露與量價微觀結構交叉比對，實時識別“申報倉位”與“實際市場行爲”之間的背離信號——將監管申報與非結構化新聞流共同轉化爲高信息密度的結構化 Alpha 信號層，實現“量價因子 + 輿情因子 + 聰明錢申報”三引擎共振研判。
Who's building this關於研究者
K2J is led by Justin, an AI quantitative systems architect with 20+ years in enterprise technology (Dell, Microsoft, IBM, private funds), now focused on production-grade agentic AI. K2J is the result — a LangGraph-based, Agentic RAG quantitative research system designed, built, and operated end-to-end in-house. K2J 由 AI 量化系統架構師 Justin 主導，他擁有 20 年以上企業級技術背景（戴爾、微軟、IBM、私募基金），專注於生產級 Agentic AI 系統。K2J 正是這一技術積累的成果——一套基於 LangGraph 與 Agentic RAG、端到端自主設計與運行的量化研究系統。
I'm also an active options and equities trader — this research isn't academic to me; I use the same regime and attribution framework in my own positions. I work in Mandarin and English, and I'm open to business partnerships and project collaborations with individuals, teams, and organizations seeking proven expertise in both AI systems engineering and quantitative market research — whether that's a joint venture, a strategic partnership, licensing, or a custom research engagement. 我同時也是一名活躍的期權與股票交易者——這套研究框架絕非紙上談兵，同樣的體制判斷與歸因邏輯直接驅動實際持倉決策。K2J 支持中英雙語，歡迎有意在 AI 量化系統與實盤市場研究領域深度合作的機構與團隊聯繫探討——包括聯合項目、戰略合作、研究成果授權或定製系統開發。
Published every trading day每個交易日更新
Complete archive of daily US equities quantitative analysis reports. 每個交易日更新一期，每日美股市場量化分析報告完整歸檔。
Frequently asked questions常見問答與技術解析
Key details on research methodology, agentic AI architecture, and collaboration. 關於量化投研框架、Agentic AI 架構設計及合作模式的核心解答。
What is K2J AI Quantitative Lab and what problem does it solve? 什麼是 K2J AI 量化實驗室？它解決了什麼核心問題？
K2J AI Quantitative Lab is an autonomous, full-market daily quantitative research system for US equities. Built to institutional standards, it solves the fragmentation between statistical macro risk modeling and actionable trade synthesis. Following each US market close, the system synthesizes macro regimes, cross-sectional breadth, 5-dimensional factor attribution, and unsupervised theme clusters into a coherent market stance and tail-risk budget.
K2J AI量化實驗室是一套面向美股全市場的每日自主量化投研系統。它以機構級標準構建，致力於解決宏觀統計風險建模與交易實盤落地之間的斷層。系統在每個交易日美股收盤後自動運行，將宏觀體制識別、截面市場廣度、五維因子統計歸因與無監督異動主題簇整合成清晰的組合策略立場與尾部風控預算。
How does the 5-layer quantitative methodology work? K2J 的五層量化多因子方法論是如何運作的？
Every issue follows a strict, orthogonal five-layer analytical pipeline: (1) Market Regime classifies Risk-On/Risk-Off macro states via GBDT ensembles; (2) Market Breadth quantifies internal market participation across volume and moving average thresholds; (3) 5-D Attribution performs robust WLS and PCA to isolate Market, Size, Value, Momentum, and Volatility return drivers; (4) Theme Clusters utilizes HDBSCAN and high-dimensional semantic embeddings to group stock movers into narrative drivers; and (5) Portfolio & Risk computes dynamic volatility targets and CVaR tail-risk constraints under Bayesian risk parity.
每一期報告都嚴格遵循自底向上的五層正交分析流水線：(1) 大盤體制識別：基於 GBDT 集成模型判別 Risk-On / Risk-Off 宏觀狀態；(2) 全市場廣度：多維度量個股實際參與度與量價背離；(3) 五維穩健歸因：使用穩健加權最小二乘（WLS）與 PCA 剝離市場、規模、價值、動量與波動率的純淨因子收益；(4) 主題簇提取：通過 HDBSCAN 與高維語義向量聚類識別異動股票背後的核心敘事；(5) 組合與動態風控：在貝葉斯風險平價框架下，輸出 CVaR 約束與倉位配置指導。
How is Agentic AI and LangGraph utilized in production? K2J 是如何利用 LangGraph 與 Agentic RAG 構建生產級 AI 架構的？
K2J runs on an autonomous multi-agent orchestration architecture built with LangGraph. Specialized agents handle automated data ingestion, statistical computation, signal cross-verification, natural language synthesis, and multilingual deployment. To ensure zero financial hallucination, the system enforces Deterministic Guardrails — mathematical proofs and hard numeric constraints that gate all generative outputs before publication. Historical market context is augmented via Agentic RAG over a Pinecone vector index.
K2J 基於 LangGraph 構建了端到端的自主多智能體（Multi-Agent）編排架構。專門的子智能體分工負責數據自動攝取、統計模型計算、信號交叉校驗、多語言自然語言生成及 CDN 部署。爲確保金融嚴謹性並徹底杜絕大模型幻覺，系統內置了確定性護欄（Deterministic Guardrails），所有輸出均需通過嚴格的數學約束檢驗；同時結合 Pinecone 向量數據庫的 Agentic RAG 進行歷史體制情境關聯。
Who is behind K2J, and what collaboration opportunities exist? K2J 團隊有什麼背景？目前支持哪些形式的商業與項目合作？
K2J is led by Justin, an AI quantitative systems architect with 20+ years of enterprise technology leadership (Dell, Microsoft, IBM, private funds) and an active options/equities trader. K2J is open to business partnerships and project collaborations — joint ventures, strategic partnerships, research licensing, or custom quantitative system development — with hedge funds, prop trading firms, asset managers, and AI ventures seeking proven end-to-end expertise in AI systems engineering and quantitative market research.
K2J 由 AI 量化系統架構師 Justin 主導，他擁有 20 年以上企業級核心技術經驗（曾任職於戴爾、微軟、IBM 及私募基金），同時也是一名美股與期權市場的活躍交易者。歡迎量化私募、對沖基金、資產管理公司及 AI 創新團隊探討商業合作——包括聯合項目、戰略合作、研究成果授權或量化系統定製開發。
How extensive is the data collected for each stock? 每期研報覆蓋多少隻股票？需要收集哪些維度的數據？
Each issue scans 1,500+ liquid US equities post-close (filtered by strict institutional liquidity thresholds to eliminate illiquid noise). For every stock, our pipeline ingests and computes across 8 core pillars (100+ granular metrics): (1) Asset Metadata & Analyst Consensus; (2) Multi-period Price, Volume & Extended-Hours Microstructure; (3) Full-period Technical Oscillators & 5-System Pivot Points; (4) Financial Statements (TTM & FQ Income, Cash Flow, Balance Sheet); (5) Advanced Cash Flow Multiples (EV/EBITDA, P/FCF); (6) Growth Rates; (7) 4 Classic Forensic Quant Models (Piotroski F, Altman Z, Sloan Ratio %, Zmijewski); and (8) Shareholder Yield. This exhaustive data network ensures our research penetrates market noise.
美股市場股票衆多，爲確保統計有效性與實盤可執行性，K2J 設立了嚴格的流動性准入門檻（過濾無成交量的微盤殭屍股），每期報告在收盤後對全市場精選的 1,500+ 只高流動性核心標的進行全量掃描與深度計算。對每隻股票計算 8 大維度超過 100 項底層數據：(1) 基礎元數據與分析師共識；(2) 多週期量價、實體意圖與盤前盤後微觀結構；(3) 全週期技術指標與 5 大流派樞軸點矩陣；(4) 財務三大報表深度數據 (TTM & FQ)；(5) 先進企業價值與現金流估值倍數；(6) 跨週期複合增速；(7) 財務比率與法務排雷四大經典量化模型（Piotroski F / Altman Z / Sloan Ratio % / Zmijewski）；(8) 股東回報與分紅政策。確保報告不是表面行情的復讀，而是穿透全市場的底層結構性歸因。
How does K2J screen the daily 20 focus stocks and what makes them unique? 研報是如何篩選每日 20 支焦點股票的？這些股票有什麼特點？
Selection follows a strict 3-layer funnel: Layer 1 (2 Extreme Movers): The market's #1 gainer and #1 decliner automatically qualify to ensure no market extreme or black swan is missed. Layer 2 (18 Resonance Leaders): Ranked by 5-D factor resonance tier (5/5★ → 4/5 → 3/5), composite attribution score, and sub-industry alignment ratio (≥50%) to filter out isolated noise. Layer 3 (Diversity Cap): Maximum 3 tickers per cluster and 3 per sub-industry to ensure macro market breadth. Screened equities exhibit strong multi-dimensional resonance where price action aligns with genuine institutional volume, cash flow support, and forensic health.
遴選嚴格遵循三層漏斗：第 1 層——極值焦點必選 (2 席)：全市場漲幅最大與跌幅最大個股自動佔位，確保極端異動或黑天鵝不被遺漏。第 2 層——5 維共振主線 (18 席)：由 5 維歸因模型根據共振星級（5/5★ → 4/5 → 3/5）、綜合歸因分與行業同向率 (≥50%) 綜合排序選出。第 3 層——多樣性約束：同一主題簇最多 3 只、同一細分行業最多 3 只，確保覆蓋市場主要產業核心主線。入選標的的共同特徵是：其異動能被宏觀、板塊、資金流、技術突破與基本面多個獨立維度同時合理解釋（共振），而非僅靠單日消息面的雜音個案。
What does the "Sector Divergence" warning badge mean? 報告中「板塊逆勢背離」預警標識代表什麼？
When a stock's price action contradicts its underlying sub-industry (e.g., gold miners rally +2.81% across 80% of peers, but one specific miner plunges -9.35%), the system flags it as Sector Divergence. The model assigns a punitive score of -2.0 to its D2 Sector dimension and deducts -0.50 from its composite Z-score, classifying it as an Isolated Divergence Drop rather than a broad short. Such drops typically stem from company-specific black swans (e.g., dilutive offerings, clinical rejections). Our AI engine extracts the exact breaking catalyst and displays an orange [Divergent] alert banner atop the card to warn against blind dip-buying.
當某隻個股的漲跌事實與所屬細分行業整體大勢發生嚴重矛盾與撕裂時（例如黃金板塊全線均漲 +2.81%、80% 個股上漲，而個股 SA 卻暴跌 -9.35%），系統會觸發「板塊逆勢背離 (Sector Divergence)」警示。模型會對其 D2 板塊共動維度賦予懲罰性負分 -2.0，綜合 Z 分扣減 -0.50，合理壓低其綜合評分，角色打標爲「逆勢背離焦點」而非「順板塊做空代表」。這類異動通常源於公司特異性突發黑天鵝（折價增發、FDA 審查受阻、業績暴雷等）。研報通過 AI 深度因果解析還原事件真相，並在卡片頂部懸掛醒目的 [逆勢背離] 預警橫幅，提醒投資者切勿盲目抄底接飛刀。
Get in touch歡迎聯繫
Open to conversations with individuals, teams, and organizations interested in business partnerships, quantitative research licensing, AI systems development, or strategic collaboration. 歡迎所有對商業合作、量化研究授權、AI 系統開發或戰略合作感興趣的個人與組織隨時聯繫。