Case Studies

Quant & AI
case studies.

001

NSE Momentum · Dual-Engine Quant Research

Factor Research Backtesting Rust
Problem

Backtests inflate easily: survivorship bias, curve-fitting, and implementation bugs make naive strategies look like alpha.

System

Survivorship-free cross-sectional momentum on NSE small/mid-caps; roughly 20 strategy families tested and rejected. Python research engine cross-validated against an independently written Rust engine, agreeing to the decimal.

Result

OOS Sharpe near 1.44, +23% CAGR, max drawdown -6.6%, Fama-MacBeth Newey-West t of about 2.7. Now running a live forward paper-trading book from a 650-stock monthly universe.

↑ 1.44 OOS Sharpe ◆ -6.6% max drawdown methodology on request
002

Kronos NSE · Real-Time Market Variance Engine

Foundation Models Real-Time Systems Market Data
Problem

Price forecasts from a single model ignore market state. Regime shifts (fear, panic, thin liquidity) silently invalidate signals.

System

Fine-tuned Kronos-small (24.7M params) on a 305K-window NSE dataset; 5-layer MVE pipeline aggregating VIX, options sentiment, FII/DII flow, GIFT Nifty, and global macro into a 5-state market classifier.

Result

Under 340ms candle-close-to-signal latency, 9-check data quality gate, and model promotion gated on out-of-sample directional accuracy beating the pretrained baseline.

↑ <340ms signal latency ◆ 139 passing tests GitHub
003

Options Pricer & Paper Trading Terminal

Derivatives Pricing Engines Risk
Problem

Retail options tooling recomputes IV from scratch every tick and ignores portfolio-level margin offsets, so realistic strategy simulation breaks down.

System

Production-grade BSM engine with full Greeks; warm-start Newton-Raphson IV solver that typically converges in a single iteration; SPAN-like holistic margin engine with 29 multi-leg strategy templates and ACID MongoDB transactions.

Result

80-90% computation reduction on IV solving in the production terminal, 66% lower API quota use, O(1) token resolution over 93K+ derivatives, and 90%+ bandwidth cut via gzip.

↑ 80-90% IV compute cut, in production ◆ 29 strategy templates GitHub
004

The Health Factory · AI Operations Suite

Production Automation ERP Integration LLM Pipelines
Problem

Two FMCG companies ran sales reporting, production planning, and supply-chain reconciliation as slow, manual, error-prone processes.

System

A portfolio of independent systems built on one pattern: deterministic pipelines over ERP and channel data, LLM document extraction behind compliance gates, and automated distribution, each shipped as a self-running service. Full architecture under NDA.

Result

Work that took days of manual effort now runs unattended on daily cycles, with delivery confirmation, retries, and audit trails throughout.

↑ 7+ systems in production ◆ architecture under NDA

Further research & builds

noarb · Option chain arbitrage checks

GitHub

Reads an option chain and reports which quotes cannot all be true at once — below intrinsic, out of order in strike, spreads dearer than they can ever pay, broken butterfly convexity, broken put–call parity. Every check is model-free, so a violation is a statement about arithmetic rather than a model's opinion, and each finding names the cheapest explanation to check first: on a real chain it is a stale quote or a missing dividend, not free money. The implied volatility solver returns nothing rather than a number where a quote does not identify one. Zero dependencies, no network, exits non-zero. MIT.

Python · 2,662 tests · 98% coverage · MIT

threeway · Reconciliation CLI

GitHub

Reconciles a purchase order against goods received against the supplier invoice and names one reason each line disagrees, covering the three ways a figure changes meaning in transit: identity, unit of measure, and period cut-off. Zero runtime dependencies and Decimal throughout, because the people who need it run it on machines where installing a library is a conversation with IT. Exits non-zero on exceptions so a pipeline can gate on a clean reconciliation. MIT.

Python · 106 tests · MIT

LLM Flash Crash

GitHub

Agent-based simulation (500 agents, 160 Monte Carlo runs) of systemic risk from homogeneous LLM traders: +30.3% drawdown amplification at full adoption. Proposed Computational Asymmetry Theory as a risk framework.

Python · Mesa ABM · Hawkes

India Quant System

Nightly ML signal pipeline for NIFTY/BANKNIFTY: 6-model ensemble (HMM, GJR-GARCH, LightGBM, LSTM, meta-model) over 80+ features, with VaR limits and a strict paper-trading gate.

Airflow · TimescaleDB · MLflow

Viveka · Equity-Research RAG

Forensic RAG engine answering investor questions over Indian listed-company filings, with citations to the exact filing, year, and page across a 250-filing corpus target.

FastAPI · Voyage · Claude

ARAS · Autonomous Research Agents

GitHub

Multi-agent pipeline orchestrating 6 AI agents from idea to finished paper; 88% retrieval accuracy over a 50K+ document corpus, 70% cut in research cycle time.

n8n · OpenAI · FAISS

DRL Quantum Sensing & Routing

GitHub

GNN-PPO routing for distributed quantum sensing networks: near-Heisenberg scaling (α=1.92±0.04) and +51% quantum Fisher information over the strongest baseline.

PyTorch · PPO · GNN

Quantum Cyclone Classification

GitHub

Hybrid quantum-classical cyclone intensity classifier using IQP-kernel QSVMs on IBTrACS storm data: 96.1% accuracy with noise-resilience and OOD analysis.

Qiskit · scikit-learn

SubPlate · Subscription Food Delivery

pnpm/TypeScript monorepo: Expo/React Native app, Fastify 5 + PostgreSQL API, shared Zod contracts typed end-to-end, and Cashfree e-mandate checkout.

React Native · Fastify · Zod

predictor-102 · Index Prediction

GitHub

Next.js financial index prediction app with real-time market data and Gemini-based forecasts; ML model over 500K+ data points with Monte Carlo simulation.

Next.js · TypeScript · Gemini