shaheer@cyber-ai — -zsh — 96×24

$ whoami

Shaheer Saud

Cyber AI Engineer // AI + Security

I build at the seam where machine learning meets security engineering — RAG/LLM pipelines, SIEM-driven threat detection, honeypots, and on-device model compression. Shipped systems at AT&T and Colgate-Palmolive; currently wrapping a CS degree at Rutgers.

ls projects/ ./contact.sh
$ cat about/profile.md

rendering markdown // 1 file

Portrait of Shaheer Saud, Cyber AI Engineer
~/about/shaheer.jpg

I'm a Cyber AI Engineer who treats security and machine learning as one discipline. On AT&T's Cyber Security Data Analytics team I architected a RAG LLM pipeline that cut document processing from weeks to seconds and pushed IAM/PAM query accuracy to 96%.

At Colgate-Palmolive I built an AI chatbot for PAM queries, ran threat hunts across Splunk, Okta, and Tanium, and automated identity migrations at scale. Across roles, my throughline is the same: turn slow, manual security work into fast, reliable, auditable automation.

Outside of work I ship offensive and defensive side projects — honeypots, data-exposure crawlers, and edge-deployed LLMs — and lead IT teams at Rutgers OIT. I care about systems that are both secure and explainable.

RAG / LLMOps Threat Detection IAM / PAM Edge ML Automation

Impact metrics

0%
IAM/PAM query accuracy on AT&T RAG pipeline
0%
faster LLM evaluation via RAGAS benchmarking
0%
help-desk workload cut by PAM AI chatbot
0%
reduction in caller wait times via automation
0%
less manual effort migrating 1,000+ IT profiles
0+
consultants led & trained at Rutgers OIT
0+
IT tickets resolved via ServiceNow
0+
labeled examples in LLM benchmarking framework
$ cat experience/*.md --reverse-chron

3 records found // most recent first

Cyber AI Engineer @ AT&T
May 2025 – Aug 2025
Middletown, NJ

Cyber Security Data Analytics Team

  • Architected AT&T RAG LLM pipeline by integrating new data sources and automating ingestion, cutting document processing from weeks to seconds and improving IAM/PAM query accuracy to 96%.
  • Developed an LLM benchmarking framework using RAGAS and Hugging Face to evaluate model configurations on 1,000+ labeled examples, cutting evaluation time by 90% and accelerating optimization.
  • Designed AI model configurations for AT&T's proprietary LLM by adjusting hyperparameters and refining embedding architectures, improving accuracy, reducing latency, and strengthening inference reliability.
Security Engineer Intern @ Colgate-Palmolive
Feb 2024 – Dec 2025
Piscataway, NJ
  • Constructed an AI chatbot for PAM queries by ingesting and curating documentation, reducing resolution time from 1 week to minutes and cutting help-desk workload by 98%.
  • Performed log analysis with Splunk, Okta, and Tanium to detect unauthorized access, anomalies, and misconfigurations, improving detection speed, strengthening response, and enhancing security posture.
  • Automated migration of over 1,000 IT profiles using PowerShell, Python, and the Colgate-Palmolive API, cutting manual effort by 80% and securely transferring credentials and secrets at scale.
Information Technology Supervisor @ Rutgers OIT
Apr 2023 – Present
New Brunswick, NJ

Office of Information Technology

  • Deployed a secure Python automation script with Selenium to streamline data aggregation across multiple platforms, reducing caller wait times by 85% and enhancing data integrity.
  • Leveraged expertise in OSI Layer 1–7 configurations, including secure IP address management, TCP/IP protocols, and DNS security, to resolve network challenges and ensure uninterrupted, secure data flow.
  • Led and trained 200 consultants, resolving over 300 IT tickets via ServiceNow, focusing on mitigating security breaches and preventing network interruptions, significantly improving incident response times.
$ ls projects/ --featured

3 featured builds // ai · security · full-stack

On-Device LLM Compression & Deployment

Advisor: Dr. Rabiul Islam

Compressed and optimized BERT/GPT-2 for edge deployment, reducing size and latency with minimal accuracy loss while deploying int8 models on Raspberry Pi and validating with LIME/SHAP.

PyTorchHugging Faceint8LIME/SHAPRaspberry Pi

Smart Home Honeypot Dashboard

Full-stack IoT deception

Built a full-stack IoT honeypot simulating a smart home, using fake login portals, Telnet exploits, and firmware traps to capture attacker behavior; visualized threats in real time via a custom dashboard.

PythonFlaskSQLiteWebSocketsGeoIPTelnetJinja2

Sensitive Data Exposure Crawler

Real-time NLP scanning

Created a real-time web crawler that scans public URLs for exposed personal data using NLP filters and visualizes scan results through a live dashboard with 100+ concurrent scanning workers.

PythonFastAPIAsyncioNLPChart.jsUvicornWebSockets
git log --more  //  additional repositories
Dictionary Attack PythonFlaskMD5 Ransomware Simulation PythonFlaskAES ScarletPlates ReactNext.jsFlaskBeautifulSoup Hand Gesture Recognition PythonOpenCVMediapipe Network Router Simulation Java Galaxy Striker UnityC#
$ cat skills/stack.json | jq

3 groups // languages · tooling · creds

Technical

~/skills/technical

PythonJavaJavaScript C / C++SQLPowerShell HTML / CSSPyTorchHugging Face

Security / IT Tools

~/skills/security

SplunkTaniumOkta IAM / PAMSIEMHoneypots TCP/IPDNSOSI 1–7

Certifications & Involvement

~/skills/credentials

Cisco CybersecurityCCNA (in progress) MSA TreasurerHackRU 2023–2025
$ cat education/degree.md

1 record

Rutgers University

The State University of New Jersey — New Brunswick

Expected May 2026

B.A. in Computer Science

gpa0.00
honorsCum Laude
trackAI & Security
$ ./contact.sh --connect

establishing secure channels...

contact.sh — secure session

$ echo "let's build something secure"

Let's connect.

Open to Cyber AI / security engineering roles and collaborations. The fastest way to reach me is email — external links below open in a new tab.

emailshaheersaud.internship@gmail.com linkedin/in/shaheer-saud github@ShaheerSaud2004 websitecipherconsulting.net websitekhafi.org downloadShaheer_Saud_Resume.pdf