About Experience Research Projects Skills Education Contact Work With Me

Ph.D. Researcher · AI Engineer · Builder

Building
Intelligent
Systems

Ph.D. candidate building production-ready agentic AI · trustworthy-AI evaluation · physics-informed geohazard ML. NSF Research Trainee 2026. Three papers accepted.

Cybersecurity Computer Vision LLM ML / DL Trustworthy AI UI / UX Data Analytics
View Projects Get in Touch
20+
AI Projects
5+
Years Research
3
Papers Accepted
Rezoan Sultan
Rezoan Sultan

About Me

AI Engineering &
Research Impact

I'm a Ph.D. candidate in Electrical & Computer Engineering at Morgan State University and an NSF Research Trainee (2026). I build production-ready agentic AI systems — spanning LLM orchestration, RAG pipelines, trustworthy-AI evaluation, and edge-cloud cybersecurity.

My research bridges the lab and production: SHAP-grounded faithfulness evaluation of LLM explanations, physics-informed landslide early-warning ML from geotechnical sensors and LiDAR, and multi-agent SaaS with tool calling and async orchestration. Three papers accepted (AGU26, TRB 2027, 7th Annual National HBCU) with more in preparation.

I've shipped 20+ AI products end-to-end — from API design and vector retrieval to monitoring dashboards and UX handoff — and I'm the founder of 'circuitcopilot.com', a free EE learning platform used by thousands of students.

🤖
Agentic LLM Systems
Planner/executor architectures with tool use, RAG, ChromaDB, and HFRL. Human-in-the-loop evals and telemetry for production reliability.
🗄️
Databases & Vector Storage
Supabase (PostgreSQL + Auth + Realtime) across production AI apps. ChromaDB for semantic vector search in RAG pipelines. SQL/MySQL for structured data.
👁️
Computer Vision Pipelines
OpenCV, YOLO, DeepSORT for real-time traffic analysis. Batched async processing with ChromaDB-powered event retrieval.
🔐
AI Cybersecurity Research
Built a unified IDS + phishing detection pipeline achieving 99.95% accuracy and 0.881 ROC-AUC. Expertise in flow-level feature engineering, adversarial ML threat modeling, and heuristic-vs-ML comparative evaluation frameworks.
🌍
Multilingual & Bengali AI
Building one of the first structured benchmarks for code-mixed Bengali instruction-following in LLMs using HuggingFace Transformers and LoRA adaptation.

Work Experience

Research &
Professional Journey

Jun – Aug 2026
U.S. Dept of Transportation
Remote / Field
Summer Research Intern
  • Developed a physics-informed instability-risk index (0–1) for rainfall-induced landslides from multi-depth soil-moisture sensors, enabling supervised ML without an observed failure event
  • Benchmarked XGBoost vs. LSTM, BiLSTM and TCN on an independent field dataset (Zhuji, China: 10 sensors, 432 records); XGBoost achieved R² = 0.70 and F1 = 0.77
  • Added a consumer-grade 940 nm LiDAR camera for periodic slope scans with M3C2 change detection, fused with the sensor-based risk index at a common decision layer
  • Lead author of two accepted papers: AGU26 poster (San Francisco, Dec 2026) and TRB 2027 Annual Meeting presentation
May 2025 – Jul 2025
Safety21 NUTC
Pittsburgh, PA
Research Intern
  • Installing geotechnical sensors (soil moisture, water pressure, weather station) at a real-world highway landslide site for real-time risk monitoring
  • Developing Python-based ML pipeline integrating real-time sensor data for early landslide detection and forecasting
  • Designing interactive dashboard to visualize sensor data, historical patterns, and ML-based risk predictions
Sep 2023 – Present
DEPA Lab · Morgan State
Baltimore, MD
Research Assistant
  • Research on agentic LLM systems (planner/executor, tool use, RAG, HFRL) and scalable backend infrastructure with human-in-the-loop evals
  • ML research spanning computer vision and time-series modeling, feature engineering, and drift monitoring
  • Teaching Assistant for Machine Learning Applications (EEGR 565) under Dr. Kofi Nyarko
May – Jul 2024
CEAMLS Summer AI
Baltimore, MD
Graduate Mentor
  • Mentored diverse undergraduate scholars from multiple disciplines on AI/CV projects
  • Provided expertise in OpenCV, YOLO, and Ultralytics for traffic violation detection accuracy and scalability
  • Supervised SQL database architecture for video data storage and management

Research & Publications

Peer-Reviewed
Research Output

✓ Accepted · Poster Presentation
Physics-Informed Hydrological Sensor-Driven Instability Prediction with ML/DL Benchmarking and Consumer-Grade LiDAR Monitoring for Rainfall-Induced Landslide Early Warning
AGU Annual Meeting (AGU26) · San Francisco, CA · Dec 10, 2026 · Session NH43C
R. Sultan (presenting author)
A two-thread prototype framework: physics-informed risk indexing from multi-depth soil-moisture sensors + consumer-grade 940 nm LiDAR with M3C2 change detection, benchmarked against LSTM, BiLSTM and TCN on an independent field dataset.
✓ Accepted · Conference Presentation
Physics-Informed Instability Risk Prediction for Rainfall-Induced Landslides: Multi-Depth Sensor Fusion with Classical and Deep Learning Benchmarking
Transportation Research Board (TRB) 2027 Annual Meeting
R. Sultan, C. Mushambo, Z. Sheng, K. Nyarko
Multi-depth sensor fusion with XGBoost, LSTM and ablation studies on sensor placement. XGBoost reached R² = 0.70 and F1 = 0.77 on an independent field dataset; mid-depth (~30 cm) sensors proved most critical.
✓ Accepted · Conference Proceedings
Fairness-Aware Credit Risk Modeling with SHAP Explainability and Demographic Fairness Audit
National HBCU Blockchain, FinTech & AI Conference · 2026
R. Sultan, M. Hossain
XGBoost credit scoring with SHAP explainability and group-aware threshold mitigation on the UCI Default of Credit Card Clients dataset. Group-aware mitigation nearly eliminated the education-based false-positive-rate disparity.
📝 In Preparation
AI-CyberCheck: Faithfulness Evaluation of LLM Explanations Grounded in SHAP Attributions
Full paper in preparation
R. Sultan et al.
Standard faithfulness metrics showed perfect scores (feature-overlap F1 = 1.0, fabrication rate = 0.0), yet 88.4–97.5% of LLM narratives exhibited explanation overreach across three models (N = 398). A new faithfulness evaluation framework for SHAP-grounded AI explanations.

Selected Projects

Production-Grade
AI Systems

01
AI CyberCheck
Unified IDS + phishing detection with reproducible end-to-end ML pipeline, model serving API, and Streamlit dashboard. Supabase backend for user auth and result storage.
99.95% IDS Accuracy · 0.881 ROC-AUC
PythonXGBoostFlaskStreamlitSupabaseSQL
↗ Live App
02
Academic Advisor AI Agent
Multi-tool HFRL-aligned LLM agent planning over student/course data. ChromaDB vector store for semantic course retrieval with human-in-the-loop offline evaluation framework.
Award-Winning · Spring 2024 Symposium
PythonLLMChromaDBHFRLSQL
⏳ In Testing GitHub Coming Soon
03
Landslide Detection ML
Sensor time-series ML pipeline ingesting soil-moisture, water-pressure, and weather streams with early warning operator dashboard. Supabase for real-time data ingestion and storage.
Real Sensor Integration · Live Alerts
PythonPandasscikit-learnSupabaseSQL
⏳ In Testing GitHub Coming Soon
04
AI Traffic Video Analyzer
CV pipeline detecting red-light running and jaywalking. ChromaDB for semantic event indexing. Batched async processing for long clips with structured SQL event storage.
Real-Time CV · Async Processing
OpenCVYOLODeepSORTChromaDBPython
GitHub Coming Soon
05
Pathology Report Analyzer
Generative NLP backend for clinical text with entity extraction, citation-backed summaries, and hallucination spot-evals. Supabase for report storage and user sessions.
RAG · Hallucination Evals · Clean API
PythonLLMsRAGOpenAISupabase
↗ Live App
06
My SupleSense
Agentic supplement recommender with policy/contraindication checks, structured outputs, and prompt-versioning. Supabase for user profiles, history, and usage metering.
Agentic · Policy-Aware · A/B Evals
PythonLLM ToolingSupabaseSQLWeb
↗ Live App
07
Businelligence AI
Multi-agent SaaS platform that automates business workflows with LLM agents using tool calling and orchestrated task hand-offs. Delivered as a freemium web product with an embeddable chat widget and messaging integrations.
Multi-Agent · Tool Calling · SaaS
ReactSupabaseLLM APIsPostgreSQL
↗ Live App
08
PharmaLens
Clinical research intelligence platform for pharma and health data teams. Turns an uploaded study dataset into LLM-generated clinical insights, automated descriptive statistics with outlier detection, and a one-click PDF report in under 60 seconds. Data processed locally — patient records never leave the client.
Claude API · Privacy-First · 60s Reports
ReactClaude APISupabaseRechartsVercel
⏳ In Testing GitHub Coming Soon
09
Circuit Copilot
Free platform helping thousands of EE and CS students understand circuits and basic electrical engineering. Features interactive circuit calculators, a hardware glossary, and curated learning paths. Built and run entirely as a public good, at no cost to learners.
Founder · Free · Thousands of Students
ReactPythonEducation AI
↗ Visit Platform

Technical Skills

Full-Stack AI
Engineering Stack

ML & Deep Learning
PyTorch / TensorFlowExpert
scikit-learn / XGBoostExpert
HuggingFace TransformersAdvanced
LoRA / QLoRA Fine-tuningIntermediate
HFRL AlignmentAdvanced
LLM & Agent Engineering
LLM Agents / RAGAdvanced
ChromaDB / FAISSAdvanced
Sentence TransformersAdvanced
OpenAI / Anthropic APIsExpert
Prompt EngineeringExpert
Databases & Backend
Supabase (PostgreSQL)Advanced
SQL / MySQLAdvanced
Python / Flask APIsExpert
Streamlit / DashboardsExpert
SHAP / Explainable AIAdvanced
Computer Vision
OpenCVAdvanced
YOLO / UltralyticsAdvanced
DeepSORTAdvanced
Data & Analytics
Pandas / NumPyExpert
TableauAdvanced
R / MATLABIntermediate
Infrastructure
GitHub Actions / CI-CDIntermediate
Linux / BashIntermediate
Arduino / Raspberry PiIntermediate
GIS & Spatial
ArcGIS ProIntermediate
Geospatial MLIntermediate
Sensor IntegrationAdvanced

Education

Academic
Foundation

Ph.D. — Electrical & Computer Engineering
Morgan State University
Baltimore, MD · Expected May 2027
In ProgressNSF Trainee 2026
M.S. — Information Technology
Washington University of Science & Technology
Alexandria, VA · June 2023
Completed
B.S. — Electrical & Electronic Engineering
American International University – Bangladesh
Dhaka, Bangladesh · December 2016
Completed
Certifications
Industry & Academic Credentials
IBM · Google · Microsoft · Morgan State
Generative AI – IBMAzure FundamentalsTableau – FinTechGoogle NetworkingIntro Data Science

Accomplishments

🏆
NSF Research Trainee — 2026
Selected as a U.S. National Science Foundation Research Trainee — a competitive program supporting graduate research excellence and professional development.
🥇
Graduate Research Poster Winner — Spring 2024
First place at Morgan State University's Spring 2024 Symposium for the HFRL-aligned Academic Advisor AI Agent, demonstrating how reinforcement learning improves student advising.
🥈
2nd Place — Morgan TechFest 2026 Tech Case Competition
Second place for CareRoute, an AI staff-matching and scheduling engine designed for a nurse-led home-health agency, competing against graduate and undergraduate teams.
✈️
Travel Award — AGU26 Conference
ED Grant Travel Support Award ($2,000) from Morgan State University for presenting research at the AGU Annual Meeting, San Francisco, December 2026. Also received full travel sponsorship for the National HBCU Blockchain, FinTech & AI Conference, 2026.

Get in Touch

Let's Build Something
Intelligent Together

Email
hello@rezoansultan.com
Phone
(443) 862-0065
Location
Parkville, MD 21234
Website
rezoansultan.com
Open To
AI Research · ML Engineering
Internships · Collaborations