data clustering image

Hello!
I'm Frederick Hayes.

Driving cutting-edge AI innovation through sustainable and human-centric solutions.

I am a highly analytical Data Scientist with over 4 years of experience in AI/ML solution development and data-driven strategy. I recently graduated with a Master of Information and Data Science from UC Berkeley (May 2025), specializing in scalable ML and Generative AI. My passion lies in leveraging data to enhance product reliability, drive informed decision-making, and create genuinely intelligent, human-centric AI solutions.

My Research Interests:
Toward Integrated Intelligence

My focus is on designing efficient AI architectures, especially neuromorphic computing and Spiking Neural Networks (SNNs), to explore novel forms of interaction with cognitive systems and the human body.

I aim to develop lifelong learning and continuous learning paradigms, leveraging Reinforcement Learning (RL), that enable seamless and intuitive human-AI collaboration and understanding. This involves exploring architectures that can help us escape the "tyranny of the Transformer," which is currently the dominant paradigm for large-scale models, to find new, more efficient, and biologically inspired ways to build AI systems. Critically, this includes investigating how linguistic symbiology, as processed by the human brain, can inform the design of AI systems that achieve greater levels of inference, adaptability, and nuanced understanding across diverse languages and cultures.

Key AI/ML Projects & Research

Here are some of my most impactful and recent projects, showcasing my expertise in developing advanced AI/ML solutions.

LAMP Primer Design Optimization

DNA Strand - Placeholder Image

Developed and deployed an AI model (1D CNN, TensorFlow) for LAMP primer design optimization. This significantly reduced false positives and enhanced the specificity and reliability of molecular diagnostic assays.

Key Skills & Tools:

  • 1D Convolutional Neural Networks (CNNs)
  • TensorFlow
  • AI Model Development & Deployment
  • Molecular Diagnostic Assay Optimization

Read the Article on Medium

NeuroBeacon (UC Berkeley Capstone Project)

NeuroBeacon Project

Developed an adaptive, RL-driven, game-based cognitive training platform empowering individuals to maintain and enhance cognitive function. It addresses uninspiring cognitive training by centering engagement and adaptability through real-time dynamic difficulty adjustment. Selected for the Spring 2025 Capstone Showcase.

Key Skills & Tools:

  • Deep Q-Learning, Pre-training (EdNet, Duolingo's SLAM)
  • Custom Reward Function, Primary/Target Networks, Replay Buffer
  • Full-stack Prototype, Cloud Deployment (AWS SageMaker, Lambda, DynamoDB, ECS, Amplify)
  • Performance Validation (simulated users, beta test data), Transfer Learning

View UCB Project Page

Retrieval-Augmented Generation for Audience-Specific Answering

RAG System

Explored the construction and optimization of a dual-audience RAG system to serve both engineering researchers and marketing professionals with tailored LLM responses, addressing a critical gap in enterprise settings.

Key Skills & Tools:

  • LangGraph-based Orchestration, Dynamic Sub-question Decomposition
  • Advanced Reranking, Hybrid Scoring, Interchangeable Models (embedding, reranker, LLM)
  • Custom Evaluation Metric (Fred Score), Hallucination Detection
  • Audience-aware Prompting, Chunk Optimization

Read the Multi-Audience Rag Final Paper Here

Rules Lawyer: NLP-Based QA System for Board Game Rulebooks

Rules Lawyer Project

Developed "Rules Lawyer," a QA system providing quick, accurate rule clarifications for complex board game rulebooks. This addressed the challenge of multi-hop queries and nuanced rule interpretation in a niche domain.

Key Skills & Tools:

  • Retrieval-Augmented Methods, Fine-tuned NLP Models (T5, Flan-T5, BERT, UL2)
  • Dataset Creation, Advanced Preprocessing, Semantic Retrieval with Reranking
  • Evaluation (ROUGE, BLEU, BERTScore F1), Qualitative Analysis
  • Hyperparameter Tuning for Rule-Intensive QA

Read the Rules Lawyer Final Paper Here

Skills & Expertise

  • AI/ML Specializations: Reinforcement Learning (RL), Natural Language Processing (NLP), Computer Vision, Generative AI, Deep Learning, Spiking Neural Networks (SNNs), Neuromorphic Computing
  • ML & AI Frameworks: TensorFlow, PyTorch, Keras, scikit-learn, NumPy, Pandas, CatBoost, LightGBM, XGBoost, Random Forest, Extra Trees, LangChain, LangGraph
  • Languages: Python, R, SQL, HTML, JavaScript, CSS
  • Data Analysis & Visualization: Predictive Modeling, Statistical Analysis, Data Visualization (Matplotlib, Tableau), Data Cleaning, Feature Engineering, Graph Analysis, Experimentation Design, A/B Testing, Neo4j
  • Cloud Platforms: AWS, Azure, GCP, Google Cloud Storage
  • Tools & Methodologies: Git, Keras Tuner, Optuna, scikit-learn pipelines, joblib, Agile, Project Management, Web Accessibility (WCAG), CMS (Cascade CMS)
  • Collaboration & Training: Technical Training, User Support, Resource Creation, Technical Writing, Cross-functional Team Collaboration

Contact Me

Have questions, or would you like to discuss a potential collaboration? Please feel free to reach out.

You can fill out the form below or connect with me via LinkedIn

My Curriculum Vitae (CV)

You can view or download a copy of my Curriculum Vitae to learn more about my professional experience, education, and research contributions.