HACKATHON SPRINT VAULT

Competitions, Sprints & Hackathons

Rapid execution, strict constraints, and working software. Here is how I approach hackathons: solving authentic problems, deploying open-weight and sovereign models, and shipping production-grade MVPs within hours.

Hacktoberfest Weekend Challenge: Build for a Friend

Briefly

Local-first legal filing assistant for automated matter classification

October 2026Completed & Shipped MVP for practicing attorney.

THE PROMPT & CONSTRAINTS

Build something with open-source AI at its core that solves a real problem for a friend.

THE SOLUTION

A local-first filing assistant built for a lawyer’s real habit: download now, file later. Watches download folders, extracts text from legal PDFs and DOCXs, and prompts an open-weight model (Gemma) to classify matters with strict confidence gating.

ARCHITECTURAL DECISIONS & TAKEAWAYS

Local Ollama chat API using compact Gemma 4 E2B with memory-safe context limits.

SQLite transaction log ensures duplicate-destination protection and atomic file moves.

Strict confidence gating: files below threshold stay safely in inbox for manual review.

Python 3.10+Ollama GemmaSQLitePyPDFpython-docxLocal-First

VIEW CODE REPO

IBM Dev Day Challenge

OnboardBob

Autonomous repository cartographer & onboarding copilot with injection defense

2026Official Challenge Submission

THE PROMPT & CONSTRAINTS

Autonomous Developer Tooling & Agent Productivity

THE SOLUTION

An autonomous codebase topology cartographer and onboarding copilot built with 3 parallel subagents (Env Auditor, Repo Mapper, Runtime Checker). Features a 3-layer prompt injection defense against poisoned repositories.

ARCHITECTURAL DECISIONS & TAKEAWAYS

Decoupled subagents running concurrent AST analyses for package graphs, environment variables, and execution paths.

Adversarial prompt sanitizer stripping hidden LLM instructions from READMEs and issue comments.

Node.jsTypeScriptAST ParsingSubagent OrchestrationPrompt Injection Defense

VIEW CODE REPO

Startup AI Buildathon / Demo Day Sprint

Research Navigator

Multi-agent business model stress-testing platform with Go proxy and judge models

July 2026Flagship Platform Deployment

THE PROMPT & CONSTRAINTS

Autonomous Multi-Agent Validation Systems

THE SOLUTION

An active validation platform that stress-tests startup business models against real-time market data, unit economics, and customer acquisition costs. Orchestrates a high-performance Go proxy runtime, Vue 3 dashboard, and Python LLM-as-a-Judge audit engine.

ARCHITECTURAL DECISIONS & TAKEAWAYS

Go proxy handles concurrent WebSocket connections and token streaming at sub-10ms latency.

Python agent engine implements a "Brutal Gatekeeper" that rejects unverified market claims.

Vue 3 Traffic Light interface gives founders instant visual confidence scores.

Go ProxyPython LLM BrainVue 3WebSocketsTavily SearchLLM-as-a-Judge

TRY LIVE SYSTEM