AI Engineer @ AsterTech | AI Materials Inc. Designer | Data Realist
Defense-focused. RL-corepilled. LLM-whisperer in the industrial fog.
- Developed an advanced CVD Anomaly Detection System with native LLM interpretation, combining dynamic thresholding, LSTM time series modeling, and structured JSON-based chat completions to explain anomalies in high-precision manufacturing systems.
- Built a full-stack multi-agent software developer platform, integrating ExaPy RAG research agents and LangChain-based autonomous code generation to architect end-to-end AI-driven development workflows. Interface built with Flask, HTML, and CSS.
- At AI Materials Inc., designed transformer-based backend infrastructure for generative synthesis of material procedures. Engineered OpenAI + SentenceTransformer pipelines, a FAISS vector store, and curated a high-quality dataset (60k→22.5k) for scientific retrieval and reasoning.
- Created an automated research pipeline for Germanium extraction — scraping, deduplicating, and processing over 10,000 open-access papers and patents. Outputs were structured into a clean, indexed dataset for LLM-ready ingestion.
- Refactored LLM interpretation systems at the Griffith Institute, grounding anomaly analysis via OpenAI’s Chat Completion API and structured context, improving clarity and interpretability across dynamic sensor environments.
- Conducted planetary systems research using Python + REBOUND simulations to analyze dynamic system responses under varied orbital conditions. (Fun fact: Jupiter is more dramatic than she lets on.)
- Embedded reinforcement learning in adversarial physics environments
- Strategic dataset hacking (manual, API, web—whatever works)
- Infrastructure scaling without sacrificing GPU sanity
- Applied prompt algebra
- How to make LLMs less annoying (a personal quest)
Interception is just intention plus inference.
If the story runs at 30 FPS, it’s already too late.