Jonathan Parran

Rebuilt IBM's CogMol conditional molecular generation framework from scratch using only the published paper—source code was never released. End-to-end pipeline: VAE molecular generation, ESM-2 protein encoding, binding prediction, CLaSS-guided sampling, and AutoDock Vina docking.

Achieved 100% molecular validity with SELFIES, 94.5% binding prediction accuracy (ROC AUC 0.97) across TMAO cardiovascular pathway targets (FMO3, CutC, CntA). Best docking score: –9.696 kcal/mol against CntA, within FDA-approved drug range.

Built and deployed an automated chromatography peak integration pipeline for LC-MS data, replacing manual LabSolutions corrections with ML-driven processing. Converts proprietary instrument files to open formats, extracts chromatogram data, and runs batch analysis.

Engineered 30+ peak-shape features and trained a multi-model ensemble (Gradient Boosting, Random Forest, Extra Trees, Ridge) achieving R² = 0.988. Processes 100+ files in minutes—an 89x speed improvement over manual analysis. Currently in production use. Featured on Cleveland Clinic Intranet Today (interviewed by Laura Greenwald).

Engineered Career Compass, a RAG chatbot for career guidance deployed on Microsoft Azure. Presented to Cleveland Clinic's AI Task Force including Chief AI Officer Dr. Ben Shahshahani. Approved for publication on the official Cleveland Clinic website.

AI-powered anti-gambling platform with 500+ users. Chrome extension with ML-based site classification blocking 200,000+ gambling sites, AI coaching, behavior tracking, analytics dashboards, and blockchain commitment vaults.

Real-time homeless assistance tool aggregating 2,500+ resources. ML-driven recommendations for shelters and pantries. Built with Eashan Kosaraju, partnered with Cleveland Metropolitan School District and Boys & Girls Club of Northeast Ohio.

AI-powered scheduling and automation platform helping local service businesses run smarter. Automated booking, client management, and intelligent scheduling for barbers, stylists, and other local professionals.

A competitive game where users battle each other for crypto. Built to explore on-chain transactions and smart contract interactions in a head-to-head setting.

Built AI models for exoplanet detection using NLP and CNN architectures on astronomical data. Mentored by Jax Rivera (MIT) and Thomas Yim (Stanford).

Applied Python to analyze biotech data and conduct computational research on SARS-CoV-2. Mentored by Dr. Manoj Samanta.

How AI systems learn to do what we don't intend. A grounded look at the gap between what we build and what we want.

The economic case for Bitcoin as hard money. Dense, opinionated, and worth it.

The real story of how Ethereum was built—the people, the chaos, and the vision behind the world computer.

A Berkeley AI professor's argument for why we need to rethink how we build intelligent systems from the ground up.

Written by a DeepMind co-founder. AI and synthetic biology arriving together—and why containment might be impossible.

A framework for thinking about crypto as an asset class before most people were paying attention.