DATA SCIENTIST & AI RESEARCHER

Sajjad Rezvani Boroujeni

Data Scientist at Actual Reality Technologies·Ohio, USA

See the workPublicationsContact

About

01 / ABOUT

I am a Data Scientist and AI Researcher working at the intersection of computer vision, deep learning, and applied statistics. I design and ship machine learning systems end to end: framing the problem, building the data and training pipeline, tuning the architecture, and carrying the result through to something that holds up in production.

My research covers generative models for data-scarce problems, segmentation and detection architectures, and the evaluation of large language models. I hold an M.Sc. in Applied Statistics and a B.Sc. in Electrical Engineering, and I publish peer-reviewed work in computer vision and applied machine learning.

  • M.Sc. Applied Statistics (Business Analytics)

    Bowling Green State University, Ohio·2023

  • B.Sc. Electrical Engineering

    Azad University of Najafabad, Esfahan·2012

Sajjad Rezvani Boroujeni

Selected work

02 / WORK
DDPM samplingscrub the timestep

Diffusion · Published research

Generative models for data-scarce detection

Detection problems that matter are usually the ones with almost no positive examples. I train denoising diffusion models to synthesize the rare class, then use the generated samples to rebalance training and lift recall where it counts.

  • DDPM
  • PyTorch
  • U-Net
  • OpenCV
Code on GitHub
Instance segmentationdrag the wipe

Segmentation · Healthcare

Medical imaging segmentation

Published work across lung cancer, brain tumor, and dental imaging. The method is consistent: U-Net variants benchmarked across CNN backbones, segmentation used upstream of CNN ensembles rather than classification alone, transfer learning from natural-image pretraining, and evaluation against clinical criteria instead of benchmark leaderboards.

  • U-Net
  • Vision Transformers
  • CNN Ensembles
  • Transfer Learning
Forecast horizondrag the now line

Time series · Control

Forecasting & process optimization

Multivariate forecasting with attributed drivers, operational regime clustering, and multi-objective control that trades efficiency against quality explicitly rather than by intuition. Extended toward digital twin simulation and reinforcement learning for real-time control.

  • LSTM
  • ARIMA
  • SHAP
  • Reinforcement Learning

Research

03 / RESEARCH

7 peer-reviewed papers across computer vision, medical imaging, and language model evaluation. 4 published, the rest under review. Expand any one for what it shows.

Full profile on Google Scholar

Experience & toolkit

04 / EXPERIENCE

Data Scientist

Nov 2024 – Present

Actual Reality Technologies

  • Lead computer vision pipelines end to end, from labeling strategy and architecture design through training, evaluation, and deployment.
  • Drive machine learning optimization for industrial process operations, improving energy efficiency and emissions while holding output quality.
  • Build multivariate time series forecasting with SHAP-based feature attribution, lag, and causality analysis.
  • Prototype multi-objective optimization, digital twin simulation, and reinforcement learning agents for real-time control.

Data Analyst

Sep 2023 – Jun 2024

AAA Club Alliance

  • Built ensemble models to improve roadside assistance efficiency across U.S. roads.
  • Led an automated variable-pay system that delivered significant operational cost savings. Recognized with the GEM Award.

Graduate Teaching Assistant

Aug 2021 – May 2023

Bowling Green State University

  • Conducted research in applied statistics and deep learning; mentored graduate students in statistical modeling.
  • Built Python frameworks for automated regression model selection and validation.

Data Analyst

Jan 2017 – Jul 2021

Araz Exir Trading Co.

  • Developed portfolio optimization and risk models (Monte Carlo, VaR) feeding real-time investment dashboards.

Deep Learning

  • Diffusion Models (DDPM)
  • U-Net
  • Vision Transformers
  • CNNs
  • YOLO
  • GANs
  • LSTM

Computer Vision

  • Object Detection
  • Instance Segmentation
  • Anomaly Detection
  • OpenCV
  • Detectron2

ML Engineering & LLM Ops

  • MLflow
  • Docker
  • Kubernetes
  • CI/CD
  • Model Monitoring
  • LangChain
  • Langfuse
  • RAG
  • Vector Databases

Languages & Data

  • Python
  • R
  • SQL
  • Spark
  • Airflow
  • Kafka
  • CUDA

Statistics & Interpretability

  • Time Series
  • Bayesian Inference
  • SHAP
  • LIME
  • Experiment Design

Recognition

  • Speaker, Great Lakes AI Week 2025, Bowling Green State University
  • GEM (Going the Extra Mile) Award, AAA Club Alliance
  • Research Collaborator, Northwest Ohio Innovation Consortium
  • Member, American Statistical Association

05 / CONTACT

Let's talk

Open to conversations about applied computer vision, industrial AI, and research collaboration.

Ohio, USA

[email protected]