Machine Learning Engineer

Paul(Qianfan) Chen

San Jose, CA Employment Authorization Cell: 979-***-**** *********@*****.*** SUMMARY

Actively seeking a position as Machine Learning Engineer or Prompt Engineer. Ph.D. graduate with a strong foundation and extensive experiences in ML and AI applications, particularly in OpenAI’s GPT-4o, Hugging Face’s Transformers, Langchain and Meta’s SAM and SD. Proficient in Java, Python, PyTorch, TensorFlow, Keras, Scikit-learn, with strong problem-solving skills. Team-oriented and result-oriented. TECHNICAL SKILLS

Programming Languages: Python, Java, C++, JavaScript, SQL, Go Machine Learning Frameworks: PyTorch, TensorFlow, Scikit-learn, Transformers, Langchain, NumPy, Pandas, SD, SAM Databases and Cloud: Chroma, LanceDB, MySQL, PostgreSQL, ElasticSearch, Google Cloud, AWS, Azure Certificate: Generative AI (Udacity), Deep Learning (Udacity), Machine Learning Specialization (Coursera) PROJECTS

Face Generation with GANs: Realistic Image Synthesis

● Designed and implemented a Deep Convolutional Generative Adversarial Network (DCGAN) to generate high-quality realistic human face images from the CelebA dataset (30,000+ pre-processed 64 64 3 RGB images).

● Built a PyTorch-based data pipeline for efficient image preprocessing and augmentation.

● Enhanced model stability with Wasserstein loss and gradient penalty, ensuring smooth convergence.

● Visualized GAN training progression with output tracking, enabling real-time adjustments and quality assurance.

● Improved image realism and diversity with advanced training techniques like generator overtraining and noise injection. Sentiment Classification with GPT-2

● Fine-tuned a GPT-2 model for binary sentiment classification on the IMDb dataset with 50,000 labeled movie reviews.

● Preprocessed data with tokenization, truncation, and padding using Hugging Face libraries to ensure compatibility.

● Achieved a 16.8% accuracy improvement (from 49.2% to 66%) through Parameter-Efficient Fine-Tuning (LoRA).

● Optimized training by targeting specific GPT-2 layers (e.g., c_attn), significantly reducing computational overhead.

● Saved and successfully loaded the fine-tuned GPT-2 model for deployment. MNIST Digit Classifier: Handwritten Digit Recognition

● Designed and trained a neural network from scratch to classify handwritten digits from the widely recognized MNIST dataset (60,000 training and 10,000 test 28 28 grayscale images).

● Preprocessed the dataset into tensors, normalize values using PyTorch and flatten the images as model input.

● Visualized the dataset to examine the size, shape, and structure of the inputs both before and after transformations.

● Built a Multi-Layer Perceptron (MLP) using PyTorch and trained it to predict the class of each input image.

● Achieved a test accuracy of 95.91% through iterative architecture refinement and hyperparameter optimization. HomeMatchAI: An AI-Powered Real Estate Agent

● Developed a Python-based application using openAI’s GPT-4o API and LangChain to deliver personalized property recommendations in natural language.

● Integrated Chroma vector database for semantic similarity searches, enabling real-time and precise property matching.

● Built a modular pipeline with prompt templates and data embedding to deliver user-focused property insights efficiently.

● Enhanced user experience by combining neighborhood insights with tailored property descriptions. AI Photo Editing with Inpainting

● Built an advanced Gradio-based app using Meta’s Segment Anything Model (SAM) and Stable Diffusion (SD) to support image editing with segmentation masks and text prompts.

● Enabled user customization of subjects or backgrounds using pre-trained Stable Diffusion XL models.

● Automated segmentation and transformation processes, delivering high-quality and realistic edits with minimal user input.

● Designed a seamless interface for effortless interaction, empowering users to explore creative photo editing. WORK EXPERIENCES

Argonne National Laboratory (Postdoctoral Researcher) Lemont, IL / Nov 2021- Oct 2023 Research Project

● Collaborated with a team of 7 researchers on quantum computation, leading the development of Python-based mathematical models to investigate the mechanisms of eNe qubits, a novel quantum computing platform. Achieved the first demonstration of eNe qubits, advancing quantum computing technology on an international scale.

● Published 20+ peer-reviewed articles with 200+ citations to date, demonstrating impactful contributions to the field and effective scientific communication.

EDUCATION

Texas A&M University, College Station, TX Aug 2017 - Aug 2021 Doctor of Philosophy, Physics

Case Western Reserve University, Cleveland, OH Aug 2014 - Jan 2017 Master of Science, Physics

University of Science and Technology of China, Hefei, China Sep 2010 - Jun 2014 Master of Science, Physics

University of Science and Technology of China, Hefei, China Sep 2005 - Jun 2009 Bachelor of Engineering, Environmental Science

Contact this candidate