Æquitas
Quant Research Platform
Connecting to API.

I'm a broke grad student, so this runs on free-tier hosting - the API is waking up from its nap and that takes a minute or two. Sorry! Play the game, read a fact, or go creep on my portfolio and resume below while you wait.

Did you know · ◈

|

Catch the tickers ◈
0 caught20s
◉Built by Adit Shah, view portfolio↗◈Hiring? View resume↗
Æquitas
v0.14.0 · Open Source

Quant research,
reimagined.

AEQUITAS is a full-stack agentic quantitative research platform that combines financial algorithms, ML models, and LLM reasoning to generate institutional-grade investment theses, automatically.

222
Unit Tests
13
Algorithms
4
Agent Nodes
13
API Routers
What it does
◈
Agentic research pipeline
A 4-node LangGraph graph autonomously researches a company, computes quantitative signals, generates an investment thesis, and self-critiques it in a revision loop.
◫
Real financial algorithms
13 production-grade algorithms: Black-Scholes + Greeks, VaR/CVaR (3 methods), mean-variance portfolio optimisation, pairs trading with Kalman filter hedge ratios.
◧
ML with explainability
XGBoost return forecaster with SHAP attribution. Hidden Markov Model regime detector classifying Bull, Bear, and High-Volatility market states.
◬
Vectorised backtesting
RSI, MACD, and Bollinger Band strategies backtested with full tearsheets: Sharpe, Sortino, Calmar, max drawdown, alpha vs buy-and-hold.
◓
Document sentiment analysis
Paste any news article, analyst report, or filing excerpt against a ticker and get LLM-scored sentiment, confidence, and key themes - not limited to what Finnhub has indexed.
◈
Draggable AI analyst
The floating chat widget can be dragged anywhere on screen - position is remembered across visits, and the panel always opens on whichever side keeps it fully visible.
Research pipeline
01
Data Ingestion
Real-time OHLCV price data from Yahoo Finance and SEC filings ingested into TimescaleDB hypertables.
02
Quantitative Signals
Momentum signals (RSI, MACD, Bollinger), pairs trading with Kalman filter hedge ratios, and regime detection via HMM.
03
ML Forecasting
XGBoost return forecaster trained with TimeSeriesSplit cross-validation. SHAP values explain every prediction.
04
Risk Engine
VaR and CVaR via three methods (historical, parametric, Monte Carlo). Black-Scholes pricer with full Greeks engine.
05
Research Agent
LangGraph 4-node pipeline: research → quant → thesis generation → critic revision loop.
06
Trade Thesis
Structured investment thesis grounded in real filings, quantitative signals, and ML evidence, with citations.
Tech stack
Agent Orchestration
LangGraph 0.2Groq LLM (gpt-oss-120b)pgvector RAG
ML & Algorithms
XGBoost + SHAPHMM Regime DetectorBlack-Scholes + Greeks
Risk & Portfolio
VaR / CVaRMonte CarloMean-Variance Optimiser
Signals
RSI / MACD / BollingerPairs TradingKalman Filter
Data Pipeline
yFinanceEDGAR APITimescaleDB Hypertables
Backend
FastAPISQLAlchemy AsyncAlembic Migrations
Frontend
Next.js 14TypeScriptRecharts
Infrastructure
Docker ComposeGitHub Actions CI/CDRailway + Vercel
Architecture
Data Layer
yFinance · EDGAR · TimescaleDB · pgvector
Agent Layer
LangGraph · Research Agent · Quant Agent · Critic Agent
Algo Layer
Signals · Pricing · Risk · Portfolio · ML Models · Backtester
API Layer
FastAPI · Pydantic v2 · SQLAlchemy Async · Celery · JWT Auth
Frontend
Next.js 14 · TypeScript · Recharts · Dark/Light theme
Built by
Adit Shah
MS Computer Science · NC State University
AI/ML · Software Development · Research · Quant Systems
Project
Versionv0.14.0
LicenseMIT
StackPython 3.13 · Node 20
Tests222 passing
CI/CDGitHub Actions
Start researching.
Ingest a ticker, run the agent pipeline,
get an institutional-grade thesis in seconds.