Ujwal Ramachandran

Hi there, I'm Ujwal! 👋

I am a

Currently pursuing Masters in Cybersecurity at NTU Singapore

Explore My Journey →

Mission Status: Active

About Me

With 3 years of experience at Optum-UHG, I've built and shipped production AI systems — from LLM-powered automation pipelines to computer vision models that drive measurable business impact.

Currently pursuing my Masters in Cybersecurity at NTU Singapore, I'm specialising in adversarial ML and AI security research — red-teaming LLMs, building multi-agent systems, and engineering intelligent defences.

My focus sits at the intersection of AI/ML engineering and AI security research — building systems that are not just intelligent, but adversarially robust.

Languages & Frameworks

  • Python
  • Shell
  • SQL
  • C
  • C++
  • CSS
  • HTML
  • JavaScript
  • Spring Boot
  • Selenium

Databases

  • MongoDB
  • Oracle DB
  • MySQL
  • GraphDB
  • Cosmos DB

AI Security & Cyber

  • Adversarial ML
  • Red Teaming
  • SIEM / SOAR
  • Zero Trust
  • Cryptography
  • Digital Forensics
  • IAM
  • Ethical Hacking
  • Keycloak
  • Wireshark
  • CTF

Cloud & DevOps

  • Azure
  • Docker
  • Jenkins
  • Kubernetes
  • CI/CD
  • Linux

AI & ML

  • LLMs
  • Agentic AI
  • RAG
  • Vector Search
  • Prompt Eng.
  • MLOps
  • XAI
  • Computer Vision
  • Reinforcement Learning
  • HuggingFace
  • Streamlit
  • YOLO v5

Bridging AI Engineering & Adversarial Research


I bring a unique blend of production AI/ML experience and security research — building, deploying, and stress-testing intelligent systems to make them both powerful and resilient.



AI & Data Science

Contract Redlining Automation

Led PoCs to automate contract redlining, track revisions, and generate insights using LLMs and rules-based logic.

Reduced legal review time and improved change accuracy

Confidence Scoring for KPIs

Created dynamic scoring metrics that adapt KPI weights based on historical and live data trends.

Enabled smarter performance decisions and accountability

Real-Time Bot Automation

Built live data pipelines that enabled bots to auto-generate orders and claims from upstream sources.

Accelerated operations and lowered manual workload

Competitive Web Scraping

Designed tools to scrape multi-site product data, bypass CAPTCHAs, and deliver reports on pricing trends.

Contributed to 10% sales growth through pricing strategy

Cybersecurity

Passwordless Authentication

Implemented secure passwordless authentication systems using Keycloak and modern security protocols for enterprise applications.

Enabled zero-trust architecture across internal apps

Application Security

Led efforts to identify and resolve security flaws through dependency checks and exploit scanning.

Reduced codebase vulnerabilities by 90%

Secure Coding Practices

Audited codebases and CI/CD pipelines for secrets, weak libraries, and insecure configurations.

Improved code quality and release readiness

OAuth 2.0 Integration

Secured enterprise web apps using Spring Boot and OAuth 2.0 for seamless authorization.

Ensured role-based access and compliance

The Power of Convergence

Where machine intelligence meets adversarial research, resilient systems emerge. My dual expertise lets me build, break, and defend AI-native systems — closing the loop between innovation and robustness.

// Featured Research

Projects & Publications

Apr 2026

AutoRedTeam-LLM

Model-agnostic red-teaming pipeline benchmarking HuggingFace instruct models across 150 adversarial prompts in 9 subcategories. Recorded an overall 81.1% ASR across 450 evaluations with zero human annotation.

Python HuggingFace Adversarial ML Streamlit 4-bit NF4
Read Blog GitHub
Nov 2025

Multi-Agent SMS Phishing Detection

Privacy-centric security system orchestrating 6 specialized AI agents to detect phishing attacks by analysing psychological triggers, URL metadata, and live web content via headless browsers.

Python LLM Selenium Multi-Agent Risk Scoring
Read Blog GitHub
Apr 2026

Agentic Network Forensics

Multi-engine, LLM-augmented pipeline for large-scale network threat investigation. Automates PCAP analysis at scale,handling gigabytes of packet captures across dozens of files that no analyst could manually review.

Python LLM Network Forensics PCAP CTF
Read Blog GitHub

System Boot Log

A chronological trace of the journey, from first compile to current mission.

2018

BTech in Information Technology

Enrolled in Information Technology at Manipal Institute of Technology, Manipal. Kickstarted a love for problem-solving, competitive programming, and early web development.

C / C++ Python Web Dev
2019

Research at IIT Delhi & Cyber Club

Research stint at Robotech Labs, IIT Delhi, built a Smart Fire Detection System with autonomous location alerts. Joined the Manipal Information Security Team, organising ethical hacking workshops and college-wide cyber events.

IoT Ethical Hacking Research
2020

First Internships & Club Leadership

Interned at Curiouz TechLab (food commodity data platform) and Virtusa Consulting Services (OAuth 2.0 + SpringBoot). Led the Agri Club as Technical & Research Head,organised IoT-in-agriculture events and maintained the club website.

OAuth 2.0 SpringBoot Data Viz Leadership
2022

Research Intern in Singapore & Graduation

Research Intern at Prognoix Pte Ltd, Singapore, engineered a passwordless login system using Keycloak, improving enterprise authentication security. Graduated BTech with 8.25 CGPA.

Keycloak Security Graduation
2022 – 2025

Data Scientist at Optum-UHG

Three years at Optum-UHG building production AI systems: LLM-powered Contract Redlining tools ($5M saved annually), proprietary Table Sense model (YOLO v5 + Azure), and enterprise security operations achieving a 90% vulnerability reduction.

LLMs Azure YOLO v5 Docker AppSec
2025

MSc Cybersecurity at NTU Singapore

Pursuing Masters in Cybersecurity at Nanyang Technological University (CGPA 4.31 / 5.0). Deep focus on adversarial ML, threat detection, cryptography, and AI security,bridging the gap between intelligent systems and resilient defence.

Adversarial ML Cryptography Threat Detection NTU
2025 – Present

AI Security Research

Building multi-agent phishing detection systems, red-teaming LLMs at scale (81.1% ASR across 450 evaluations), and conducting agentic network forensics,applying AI to harden and challenge modern security boundaries.

Red Teaming Multi-Agent AI LLM Security Active

Ask Me About

Beyond the code and algorithms, I'm passionate about exploring diverse fields that fuel my curiosity and creativity.

AI / ML

Intelligent systems, automation, and the transformative potential of machine learning and LLMs.

Cybersecurity

Threat analysis, defense strategies, and building resilient security architectures for the digital age.

Astronomy

Space exploration, celestial phenomena, and the wonders of the universe that spark scientific thinking.

Photography

Wildlife and nature photography, capturing the beauty of the world through a creative lens.

Establish Uplink

Whether you're interested in AI innovation or just want to chat.

Let's Connect

Chat over Email
My LinkedIn Profile
My GitHub
Warning You are not Active enough.