Agentic AI Engineer · AI Security Researcher · Data Scientist

I build things at the intersection of Agentic AI and Cybersecurity - multi-agent systems, LLM security pipelines, and adversarial ML, grounded in 3 years of production Data Science at Optum - UHG and an ongoing MSc in Cyber Security at NTU Singapore (CGPA: 4.31/5.0).

My work spans both sides of the AI-security boundary: shipping LLM-powered tools in enterprise environments and stress-testing those same systems for vulnerabilities.

My Portfolio Website is more than just a list of skills, it's an interactive journey through my digital mind.

While it maps out my professional evolution across Data Science and Cybersecurity, it also rewards the curious. Venture off the main path to discover a whimsical Pixel Art World populated by a crew of digital animals, or test your wits against a Hidden Logic Game concealed deep within the system.

(Hint: The footer might hold the key to the firewall...)

• 🎓 MSc in Cyber Security at NTU Singapore (Aug 2025 - Present)
• 🔐 Coursework in Security Monitoring, Adversarial ML, Privacy Preserving Technologies, and Cryptography
• 🤖 Building agentic AI systems applied to real security problems
• 📡 Researching LLM red-teaming, prompt injection, and multi-agent orchestration

skills = {
    "Agentic AI & LLMs": [
        "Multi-Agent Systems", "LLM Integration", "RAG", "Prompt Engineering",
        "Vector Search", "Hallucination Mitigation", "Tool Use"
    ],
    "AI & ML": [
        "Adversarial ML", "MLOps", "Deep Learning", "Explainable AI (XAI)",
        "Computer Vision (YOLO v5, OpenCV)", "Reinforcement Learning"
    ],
    "Security": [
        "LLM Red-Teaming", "Application Security", "Zero Trust Architecture",
        "CI/CD Security (Prisma, Azure Defender)", "IAM (Keycloak)", "SIEM/SOAR",
        "Digital Forensics", "Cryptography"
    ],
    "Languages": ["Python (Pandas, NumPy, Scikit-learn)", "SQL", "Shell", "C", "C++"],
    "Infrastructure": [
        "Azure (AKS, Cosmos DB, Document Intelligence)", "Docker",
        "Kubernetes", "Jenkins", "MongoDB Atlas", "Oracle DB", "Linux"
    ]
}

• 🛡️ AutoRedTeam-LLM: Model-agnostic red-teaming pipeline benchmarking LLM safety across 150 adversarial prompts. Achieved 81.1% ASR across 450 evaluations with zero human annotation, using a two-stage semantic classifier and 4-bit quantized local inference.
• 🕵️ Scam Detection Agent: Privacy-centric multi-agent system orchestrating 6 specialized AI agents to detect SMS phishing via psychological trigger analysis, URL metadata, and live web content. Engineered Context Object pattern to decouple agents and mitigate LLM hallucinations.
• 📄 Contract Redlining Tool: Built end-to-end LLM-powered document comparison tool at Optum with collaborative section-based access, reducing manual processing by 45% and saving $5M annually vs third-party vendors.
• 🔍 RAG Auto-Suggestion System: Engineered retrieval-augmented generation pipeline for drug language and rebate exclusion suggestions using vector search and prompt engineering.
• 📊 Confidence Score Metric: Designed proprietary KPI weighting system with dynamic historical data adaptation, reducing contract review time by 70% and delivering $3M in annual savings.
• 🕷️ Web Scraping Framework: Built Python + Docker competitive price tracking system with CAPTCHA bypass, contributing to a 10% sales increase.

• 🔴 Agentic Network Forensics: Agentic AI system for automated network traffic analysis and threat investigation.
• 🔍 Security Automation at Optum: Achieved 90% vulnerability reduction through automated CI/CD scanning using Prisma and Azure Defender; screened exploitable libraries and secrets across production deployments.
• 🔐 Passwordless Authentication: Engineered enterprise passwordless login using Keycloak, enabling Zero Trust architecture across internal applications.
• ⚙️ Secure SDLC: Audited codebases and CI/CD pipelines for secrets, weak libraries, and insecure configurations, improving code quality and release integrity.
• 🛠️ OAuth 2.0 Integration: Secured enterprise web apps using Spring Boot and OAuth 2.0 for seamless role-based authorization.

• 🪐 Astronomy - My passion for space and celestial phenomena.
• 📷 Photography - Capturing moments and wildlife through my lens.
• 🖥️ AI Security - LLM red-teaming, adversarial ML, and defense strategies.
• 🤖 Agentic AI - Multi-agent systems, orchestration patterns, and production LLM pipelines.

• 📸 Photography Enthusiast: Wildlife and nature photography (not professional though).
• 🌱 Environmentalist: Active volunteer for sustainability initiatives.
• 🎮 Games: Whether outdoor sports or online video games, I'm in.
• 🎓 Mentor: Helped underprivileged children through educational programs.

• 📧 Email: contactme.ujwal@gmail.com
• 💼 LinkedIn: Ujwal Ramachandran
• 🌐 Portfolio: ujwal-ramachandran.github.io

⭐️ From Ujwal-Ramachandran

"Building at the intersection of AI and security : Where the most interesting problems live."