Hi, I'm AJ, an aspiring AI Research Engineer in Embodied AI & World Models. enjoy writing software programs that I have a vision for, for example: Immersia & Syneuro both of which are currently under development. I also enjoy independent research where I can write papers, perform experiments, and analyze subjects in AI: Machine Consciousness & Deep Learning.
To sum up, I'm primarily driven by the why and how, where I often look for answers theoretically and practically.
- Backend Architecture [Python]: Microservices, Event-driven
- Systems & Low-Level Programming - C++, Operating Systems
🔬 AI Research & Experimentation
• First-Token Broadcasters: Mechanistic Origins of Language Identity and Distributed Robustness in Transformers: Paper on the mechanistic origins of language identity and distributed robustness in Transformers.
• FTB Graph: Determining and Validating First-token Broadcasters and Language-Identity Head Circuits in Multilingual Language Models: Paper on determining and validating first-token broadcasters and language-identity head circuits in multilingual LLMs.
🌌 Immersive Worlds & Conversational Agents
• Immersia [WIP]: Generative-AI visual novel immersive world.
• Syneuro [WIP]: Neuro-symbolic affective-computing framework for conversational agents.
⚡ Systems Design & Infrastructure
• ParamFinder: A system architecture proposal for a cloud-agnostic orchestration layer for asynchronous hyperparameter optimization.
• RailDB [WIP]: A personal database built in C++, inspired by RDBMS architecture.
• QMonitor: Python monitor tool for quantum experiments in AWS and local environments.
• QuantumRNG: Quantum random number generator simulation built with Qiskit.
🌐 Applications & Web APIs
• Waitlister API: A lightweight TypeScript API for product waitlist signups and referrals.
• Grayscale-Filter: A mini application built with OpenCV for grayscale filtering.
🤖 Applied Machine Learning & Deep Learning (Mini Projects)
• feedforward-neural-network: A Feedforward Neural Network built from scratch using PyTorch.
• ZombieApoc-SupplyRun-Classifier: A mini XGBoost classifier model trained for status screening.
• agriculture-prediction-model: A precision agriculture script identifying predictive soil nutrients.
• anime-ratings-prediction-model: A ML pipeline leveraging CatBoost Regression to predict ratings.
• reinforcement-learning-exercise: A foundational RL-inspired Python script.
🤝 Open-Source & Community Contributions
• BIP324-MITM: Security contribution exploring Bitcoin's P2P encrypted transport protocol man-in-the-middle vectors.
• [jetspace-imitation-learning]:
- LeeMarshall1113/jetspace-imitation-learning#9 - Optimized dataset loading by implementing options-based selective NPZ decompression and adding fail-first guard clauses to skip invalid episodes early.
- LeeMarshall1113/jetspace-imitation-learning#11 - Refactored data pipeline architecture by decoupling Episode into a standalone dataclass, upgrading EpisodeBuffer storage mechanics, and updating the demo collection script for compatibility.
• Room-TBA: Contributed a defensive data-fetching and caching strategy to the Progressive Web App (PWA).
📝 Technical Writing
• A Mini Blog in Neuromophic Computing Security: Explanations and insights exploring neuromorphic computing architectures.
- Reading theoretical whitepapers & engineering blogs
