I am a Senior Software Engineer focused on designing backend systems that are scalable, fault-tolerant, and maintainable.

My engineering work centers around Java, Spring Boot, distributed systems, event-driven architectures, APIs, caching, persistence, and system design.

I am especially interested in the engineering problems behind production systems:

• service reliability under failure,
• asynchronous communication,
• event-driven processing,
• latency and caching,
• API and data-layer design,
• and security boundaries in increasingly AI-enabled applications.

More recently, I have been exploring program analysis and AI security, particularly how traditional software-security techniques can help protect applications interacting with large language models.

Policy-Aware Static Taint Analysis for LLM Data Security

PIIFlow is a Python-based static analysis tool for detecting potentially unsafe flows of sensitive information into LLM APIs.

It brings traditional program-analysis techniques into a modern AI-security setting by analyzing how sensitive data propagates through application code before reaching an external model endpoint.

Sensitive Source
      ↓
Application Code
      ↓
Taint Propagation
      ↓
Policy Evaluation
      ↓
LLM Sink
      ↓
Security Finding

Core Areas

Python · Static Analysis · AST Analysis · Taint Tracking · Data-Flow Analysis · DevSecOps · AI Security

Problem

Applications increasingly send user, business, and operational data to LLM APIs. Traditional runtime monitoring may detect a problem only after information has already left the application boundary.

Approach

PIIFlow explores whether potentially unsafe flows can be identified before execution by analyzing program structure and propagating sensitivity information through source code.

Core idea

source → propagation → policy → LLM sink → finding

Explore PIIFlow →

Backend Architecture · Distributed Systems · Event-Driven Systems · System Design · Performance · AI Security

I am particularly interested in systems that must remain reliable under scale, resilient under failure, and understandable as complexity grows.

My current technical direction also explores the boundary between traditional backend engineering and emerging AI systems — especially where ideas from program analysis, security, observability, and distributed systems can improve reliability.