I have been working in LLM System development since the release of the ChatGPT API. Since then, I have been Head of AI at two different companies, completed numerous contracts, and started an AI Research company where we explore novel approaches and algorithms for LLM systems. I have built up a set of custom tools and solutions for many of the common problems, like hallucinations, scaling, reinforcement learning, RAG, and quality evaluation.
My strong background in mathematics and AI gives me a much deeper understanding of how these systems work, and as a former Computer Science educator, I'm able to effectively communicate the possibilities – and limitations – of LLM systems. I excel at bridging the gap between complex new technologies and product design.
I am currently looking for additional projects in the following areas:
1. LLM System Design
a. Top-level design (agentic or
b. Product Viability
c. Practical cost analysis
d. Existing Codebase Evaluation
e. Proof-of-concept and MVP building
f. Infrastructure analysis and design
2. Custom algorithm design/development
a. Advanced semantic similarity/NLP systems
b. Knowledge graphs
c. Other complex data structures + algorithms
3. Fast web app development
Using custom coding tools that I have built using LLM systems, I can build robust, reliable web applications extremely quickly, with test cases, understandable code, CI/CD, security, etc.
If you want a good solution fast, I can help with that. Preferred stack is flask/psql/celery/redis/docker or Django.
Current Role:
Head of AI at Cadenza. Owner/researcher Guru Cloud & AI LLC
Current Research:
Dynamic Embedding-based Categorizational Trees
Self-Contained Dynamic Agentic LLM systems
Relational Semantic Embeddings
Dynamic, Task-Relevant, Computationally-curated Memory
Automatic Software Dev Tools
Previous LLM projects:
Cadenza: agentic system with graphics, research, high-volume RAG, artifact-based report generation, api learning system that uses reinforcement learning to learn to use any API from documentation, dockerized and scalable multithreaded environment.
NewSeat: Dynamic in-conversation matching between candidates and jobs using a dynamic categorical tree, RAG on internal documents, in-conversation job applications, genai tools for creating job listings, application management.
Lula, a solar expert who guides customers through the quoting process for a solar system and stores info it learns in salesforce.
Jimmy, an AI who texts and qualifies people for healthcare. The conversations are automatically logged, tracked, analyzed, and maintained.
Components: Python, Google Cloud, Twilio, Notion (for user interface), OpenAI. Completely serverless, runs at 4c/conversation and can manage nearly infinite scaling.
Many more projects viewable on my Github and website (site in progress, see sidebar), but I have experience with all of the following:
Java, Python, C#, Visual Basic, Unity
CSS, javascript, PHP, html, nodeJS, .Net
Google Cloud, AWS, Azure
OpenAI, Anthropic, Llama, etc.
PyTorch, DNN, model training/fine tuning (and why you probably shouldn’t be doing that in most cases)
PromptFlow, LangChain (and why we can do better)
SOAP/REST, flask, ngrok, nginx, django
JWT, Oauth, https, ssh, general web sec
Fullstack dev w/ docker, celery, docker-compose, CI/CD
Google Sheets, Excel automation, Selenium/browser automation
CRM systems (Salesforce, etc), API integrations
SQL, PSQL, SQLAlchemy, SQLLite, no-sql
Data analysis
PDE Modeling, Numerical Methods, genetic algorithms, Neuron chain models
Learned Scoring Algorithms
AI-Generated Code Specialists
