Benji
💙

Hello, I'm Benji Peng. I started my first research project 11 Years ago. I have gained experience in applying automation and ML to address complex biological and chemical problems. My focus is on Machine Learning, Product Development, and Biophysics, and statistical modeling for cutting-edge research.

Contact Me Here Download CV

About Me

I grew up helping my mother with her e-business for pet supplies, which naturally made me tech-forward and gave me an early understanding of digital workflows and customer-facing technology.

I come from a family passionate about life sciences, with two cousins holding Ph.D. in the related fields. That love for science is what motivated me to pursue my education, with an emphasis on applying computational techniques to solve complex problems.

After getting my Ph.D. candidacy, I pursued my passion for machine learning to enhance my research capabilities. My current work is mostly project-based, focus bring novel technology to the broader audiences.

Outside of my work, I enjoy traveling ✈ī¸, art đŸ–ŧī¸, techno đŸĒŠ, and spending time with my dog đŸļ.

My Research

Securing Large Language Models: Addressing Bias, Misinformation, and Prompt Attacks

This paper reviews security concerns in large language models, focusing on bias, misinformation, detection mechanisms, and vulnerabilities like prompt attacks.

  • LLMs
  • Security
  • Bias
  • Misinformation
  • Prompt Attacks
  • Detection
  • Mitigation

Large Language Models and Cognitive Science: A Comprehensive Review of Similarities, Differences, and Challenges

This review explores how Large Language Models (LLMs) relate to cognitive science, discussing their similarities to human cognition, evaluation methods, and potential applications. It addresses cognitive biases, limitations of LLMs, and suggests methods for enhancing their performance.

  • LLMs
  • Cognitive Science
  • Evaluation
  • Applications
  • Cognitive Biases
  • Artificial Intelligence

Emerging Techniques in Vision-Based Human Posture Detection: Machine Learning Methods and Applications

This article reviews machine learning methods for human posture detection, covering 2D and 3D detection methods, image and video-based approaches, and the use of open datasets for training. It explores applications in healthcare, surveillance, and human-computer interaction.

  • Human Posture Detection
  • Machine Learning
  • 2D Detection
  • 3D Detection
  • Healthcare
  • Surveillance
  • HCI

Optically Modulated and Delayed Fluorescent Proteins for Sensitive Cellular Imaging

This study introduces optically modulated and activated delayed fluorescent proteins for biological imaging. It details the engineering of YFP variants to create new emitters with unique modulation profiles and delayed fluorescence, enhancing sensitivity in cellular detection.

  • Fluorescent Proteins
  • YFP
  • Cellular Imaging
  • Delayed Fluorescence
  • Optical Modulation
  • Biological Detection

Measuring Unbiased Retention Factor vs. Temperature Relationships in Gas Chromatography

This paper introduces a methodology for accurately measuring gas chromatographic retention factor vs. temperature (k vs. T) relationships, overcoming the bias caused by non-ideal GC system behavior. It improves compound identification accuracy by leveraging back-calculated temperature profiles and free software for easy application.

  • Retention Factor
  • GC-MS
  • Retention Projection
  • Chromatography
  • Temperature Programs
  • Compound Identification

My Skills

My Experience

AppCubic

Research Scientist & Technical Consultant

Miami, FL (Remote)

At AppCubic, I designed and implemented ML-driven digital infrastructures and worked closely with researchers worldwide on advanced machine learning projects. I co-authored multiple research articles, including work with NIH entrepreneurship partners. My key project, TradeGPT, involved the development of an LSTM and reinforcement learning model for equity and digital asset analysis, integrating language models for named entity recognition tasks. This research-driven approach helped attract institutional investors and led to licensing opportunities for the model.

2018 – Present

Kaseya

Software Development Engineer

Miami, FL

As part of Kaseya's leadership development program, I spearheaded the creation of a comprehensive API testing framework, reducing manual QA intervention by over 82%. My research-driven mindset led me to advocate for type safety and data cleanliness in network topology discovery, ensuring scalability and robustness. Additionally, I mentored team members in various testing methodologies, including AI-assisted testing and visual regression testing, which further honed my skills in analyzing and improving software performance.

2022 – 2023

Georgia Institute of Technology

Ph.D. Candidate in Physical Chemistry / Biophysics

Atlanta, GA

My doctoral research focused on applying unique statistical binning methods to scatter patterns in flow cytometry data for antibacterial susceptibility tests. I integrated machine learning algorithms to improve accuracy and predictive power, and used high-throughput pipelines for biomarker design and tissue imaging. Throughout my Ph.D., I employed 3D visualization tools and mathematical modeling, which were implemented using test-driven development principles (95% coverage). I also conducted customer discovery sessions with lab managers and pathologists to tailor our solution for real-world applications.

2017 – 2021

University of Minnesota

Undergraduate Research Scientist / Junior Scientist

Minneapolis, MN

During my time as an undergraduate researcher, I worked on gas chromatography automation and retention time prediction for building a plant metabolite library. I later contributed to an open-source project led by my advisor, where we wrote Java code for compound prediction based on retention times. I presented my research at multiple conferences and published my first scientific paper, honing my skills in data analysis, scientific communication, and coding for research applications.

2012 – 2014

Contact Me