Introduction
I’ve recently conducted an artificial intelligence and machine learning deep dive, using extensive hands-on coding and hackathon competitions as a way to create tangible implementations and connect with others in the industry. Over the course of this deep dive, I’ve been able to compete in a broad spectrum of competitions, tally a number of wins, and accumulate more than $47,000 in GPU and API prize credits in a short time.
AI and ML Hackathons and Projects
Chess AI Hackathon Series (Six-Time Winner)
August 2024–March 2025
Formed and led a team of three in competition against ten San Francisco and Sydney teams to build the strongest possible chess-playing deep learning model in only two days. Sponsored by Strong Compute (YC W22).
Explored two high-level approaches: a vision-like CNN model and a language-like GPT model. A key inspiration for the GPT-based models was Adam Karvonen’s paper Emergent World Models and Latent Variable Estimation in Chess-Playing Language Models. Within the vision approach, we experimented with CNNs, dilated convolution, ResNets, transformers, and multi-head self-attention. Trained on a 72xA100 cluster. Won the first tournament with a record of 8 wins, 1 loss, 1 draw, earning a $10K GPU compute credit. Released model source code and a case study.
Returned as a solo entrant to win five out of six subsequent instances of the tournament.
Chess Theme and Opening Multilabel Classifier (Grant Recipient)
April 2025–July 2025
Research project to implement a deep convolutional neural network to perform multi-label classification of board positions sourced from the lichess puzzles dataset. Each board position is labeled with applicable themes (for example, back rank mate, zugzwang, advanced pawn, etc.) as well as openings, if relevant (for example, Sicilian Defense, The English Opening, etc.).
A model is trained to recognize the positions and themes and then to apply the same classification mechanism to positions during live games or during review of unseen games.
ARC-AGI Hackathon ($10K Grant Recipient)
September 2024
Formed a team of three to develop and pitch an approach to the $1M ARC-AGI Challenge (Abstraction and Reasoning Corpus for Artificial General Intelligence). Explored CNNs, Autoencoders, language models, and program generation. Investigated ARC-DSL domain specific language. Ultimately proposed use of the Peano formal theorem proving environment. Awarded a $10K research grant.
DeepTruthSeek (Research Project)
February 2025
Project to create a DeepSeek-r1-Llama-70B fine tune to improve detection and description of logical fallacies in written text. Used the CoCoLoFa dataset from “CoCoLoFa: A Dataset of News Comments with Common Logical Fallacies Written by LLM-Assisted Crowds” and PyTorch Fully Sharded Data Parallel (FSDP) training.
Anthropic Claude Plays Pokémon Hackathon (Grand Prize Winner)
March 2025
Formed a team and competed against 120+ hackers to create the best Pokémon playing agent based on Claude 3.7 Sonnet, determined by distance-from-origin progress through a notoriously challenging level as the objective function. Experimented with agent prompt engineering, vision coordinate overlays, splitting/sub-specialization of tools, and alternative context compression algorithms.
Multimodal/Video Hackathon, OIX and Susa Ventures (Best Technical Execution)
April 2025
Created a realtime AI-enabled video avatar for conversational assistance in scheduling and other administrative tasks. Implemented a solution using ByteDance LatentSync 1.5 for video/audio lip synchronization, Sesame CSM (Conversational Speech Model) for realistic text-to-speech, Google Calendar API for scheduling and the WebAudio API and Mavi Transcription APIs for capturing user questions for the agent via browser microphone input.
Y Combinator/Anthropic World’s Largest MCP Hackathon
May 2025
Built a project which connects Claude Desktop to a locally hosted Blender MCP server, a local Blender instance, and a newly-developed local MCP server which serves as a technical drawing database and server for USPTO technical drawing.
The user can name a historical engineering mechanism or artifact. For example, the Wright Brothers Flyer, a planetary gearbox, etc. Claude then generates a 3D model directly from the 2D technical drawings. The user can modify the model by changing parameters, materials, etc., and can ask for information about the model’s mechanism, function, and effects of any modifications.
Nous Research Atropos RL Environments Hackathon (Distributed RL)
May 2025
Created an environment for distributed reinforcement learning within the Nous Research Atropos framework. Generated a set of 1K conversational samples with (preferred, dispreferred) response pairs via a meta-prompt. Used these response pairs for DPO training of conversational models to align with individual user preferences in tone, politeness, etc.
Professional Certifications
AI & LLM Development
| Certification | Issuer | Date | |
|---|---|---|---|
| Google Cloud Attention Mechanism | Dec 2025 | Credential | |
| Google Cloud Introduction to Image Generation | Dec 2025 | Credential | |
| Introduction to Model Context Protocol | Dec 2025 | Credential | |
| Building with the Claude API | Nov 2025 | Credential | |
| Claude Code in Action | Nov 2025 | Credential | |
| LangChain Essentials – Python | Nov 2025 | Credential | |
| LangGraph Essentials – Python | Nov 2025 | Credential | |
| Neural Networks and Deep Learning | Oct 2020 | Credential |
Product & Process
| Certification | Issuer | Date | |
|---|---|---|---|
| Inspired: How to Create Tech Products Customers Love | Apr 2015 | ||
| Certified ScrumMaster | Oct 2008 |
Professional Skills
AI & Machine Learning
| Skill | Description |
|---|---|
| Attention Mechanism | Transformer attention layers, self-attention, cross-attention, sequence modeling, scaling laws, modern LLM architectures |
| Diffusion Models | Diffusion-based generative modeling, DDPMs, latent diffusion, text-to-image systems |
| Large Language Models | Architecture, training paradigms, inference optimization, evaluation, deployment of transformer-based models |
| Transformers | Encoder-decoder models, self-attention scaling, positional encoding, pretraining/fine-tuning dynamics |
| Deep Learning | CNNs, RNNs, optimization, regularization, representation learning, large-scale training |
| Machine Learning | Supervised, unsupervised, reinforcement learning; model evaluation; feature engineering; production ML systems |
AI Agents & Orchestration
| Skill | Description |
|---|---|
| AI Agents | Autonomous/semi-autonomous LLM agents, tool use, planning, memory, state machines, human-in-the-loop workflows |
| Coding Agents | LLM-driven agents for repo navigation, refactoring, cross-file reasoning, test generation, code transformation |
| Model Context Protocol | MCP servers/clients for secure, structured tool and data access for LLM agents |
| LangChain | Chains, agents, tools, retrievers, memory, LangSmith observability, modular LLM application design |
| LangGraph | Graph-based control flow for multi-actor, cyclic, long-running LLM workflows with durable state |
| Prompt Engineering | Structured prompts, tool schemas, conversation protocols for reliability, controllability, reasoning depth |
| RAG | End-to-end pipelines: embedding, chunking, vector search, hybrid retrieval, grounding in knowledge bases |
Engineering & Systems
| Skill | Description |
|---|---|
| Software Engineering | End-to-end lifecycle: architecture, implementation, testing, deployment, operations (GitHub, CI/CD, Unix) |
| Distributed Systems | Scalable, fault-tolerant architectures, data pipelines, large-scale backend systems |
| Cloud Applications | Cloud-native, high-throughput, globally distributed services (Google Cloud, BitTorrent) |
| Data & Analytics | Metrics systems, analytical pipelines, executive dashboards (SQL, Big Data, Data Visualization) |
Web3 & Product
| Skill | Description |
|---|---|
| Web3 / Blockchain | Ethereum ecosystem, wallets, smart contracts, DAOs, zero-knowledge systems (Consensys, MetaMask) |
| Product Management | Product strategy, OKRs, roadmapping, execution, growth, cross-functional alignment |
| Developer Relations | API platforms, SDKs, community growth, technical evangelism, developer experience (MetaMask) |