AI-Powered Security API Gateway

A stateless, high-throughput security gateway that detects and mitigates malicious traffic in real time using an event-driven architecture, enforcing gateway-level protection before requests reach backend services.

Key Design Decisions:

• Problem → Detect and block malicious clients without increasing request-path latency or consuming backend resources
• Solution → Fire-and-forget Kafka event pipeline with Redis-backed blacklist enforcement, Strategy Pattern-based threat detection, and AI-assisted analysis
• Consistency → Identified and resolved a race condition between Kafka consumer ordering and Redis cache writes using a write-through caching strategy

Highlights:

• 🪪 Cryptographically signed Visitor ID tracking for secure client identification
• ⚡ Kafka-based asynchronous threat analysis with fully decoupled security services
• 🎯 Strategy Pattern-driven detection engine for rate abuse, bot activity, and behavioral anomalies
• 🔄 Circuit Breaker optimization to terminate analysis immediately after threat detection
• 🚫 Dynamic Redis-backed blacklisting with configurable time-based enforcement
• 🤖 Ollama/Gemma 3 integration for AI-assisted threat classification when rule-based strategies fail
• 📊 Centralized security logging, SOC dashboard (React/TypeScript), and historical context analysis

Consent-Driven Distributed Identity Platform Alpha 🚧

A microservices-based KYC orchestration system that eliminates redundant verification across institutions while enforcing secure, consent-based data access.

Key Design Decisions:

• Single Source of Truth → One verified identity reusable across organizations
• Consent → Explicit user authorization required before any data access
• Event-Driven → Asynchronous communication using Kafka
• Security → JWT authentication, OTP-based DigiLocker verification, and encrypted data storage
• Lifecycle → Automated renewal, revocation, and update workflows

Highlights:

• 🔐 Consent-based identity sharing with strict access control
• ⚡ Event-driven microservices using Apache Kafka
• 🏛️ Multi-institution KYC orchestration
• 📊 Strong data governance and trust boundary enforcement

Automated Threat Detection Pipeline

Static analysis + ML pipeline to classify polymorphic malicious payloads hidden within PDF structures.

Key Design Decisions:

• Innovation → Custom structural parsers to identify high-entropy object streams in malformed PDFs
• Accuracy → XGBoost ensemble model with SHAP-based interpretability for classifying zero-day polymorphic variants
• Usability → Streamlit interface for quick, interactive scanning

Highlights:

• 🧠 XGBoost classification pipeline with SHAP explainability
• 🔬 Custom PDF structure parser for malformed documents
• 📈 High-entropy stream detection

Systematic Algorithmic Trading Engine

An autonomous intraday/swing trading system for NSE/BSE that scores every candidate trade on a weighted, multi-factor model and enforces a hard risk gate before any order goes live.

Key Design Decisions:

• Signal Fusion → Composite scoring: 30% technical (EMA crossover) · 30% fundamental · 20% sentiment · 20% event-driven
• Risk Management → Hard-coded risk gate — 1.5% per-trade stop loss, 3% daily circuit breaker
• Execution → OpenAlgo/Zerodha integration for live order routing and TradingView-based charting

Highlights:

• 🧠 FinBERT-based sentiment analysis on market-moving news
• 📊 EMA crossover + composite multi-factor signal generation
• 🛑 Automated circuit breaker for capital protection
• 📉 Live order execution via Zerodha with real-time charting