Applied AI | Autonomous Research | Intelligence Systems | Systems Engineering
I build research and operational systems that turn scattered information into evidence people can inspect and use. My work connects applied AI, financial research, infrastructure and automation, shaped by experience with real equipment, incomplete information and practical constraints.
The aim is consistent: investigate a useful question, make the reasoning visible and give the next action a sound basis.
Information becomes intelligence when it improves decision quality.
Building intelligence that earns its conclusions.
ATM investigates whether autonomous research systems can identify market opportunities, challenge assumptions and produce independently evaluated economic evidence. The private architecture emphasizes falsifiable hypotheses, research governance, deterministic evaluation and risk boundaries.
The public research core provides event studies and episode summaries with strict input validation, explicit costs and reproducibility hashes. It is one limited component of ATM, with no provider, inference or execution connection. Repeatable financial profitability remains unproven.
Explore the research core · Architecture and boundaries
Turning scattered information into actionable intelligence.
FIELDNOTE helps an independent researcher investigate people and institutions, connect evidence, manage correspondence and coordinate action. Questions, source records, draft messages and next steps stay connected, while reviewed public updates remain separate from private notes.
The public synthetic demonstration uses invented records to show the workflow, strict data contracts and recoverable local revisions. Clean installation, seven automated checks and 11 browser/service observations are recorded. The private production installation is not distributed.
Explore FIELDNOTE · Architecture, setup and workflow
An autonomous racing controller combining track telemetry, control design and measured qualification. Published evidence records an 84.388-second Corkscrew standing-start lap with zero damage, repeated timing qualification and a GUI reproduction. Source, offline checks and reproduction instructions accompany the result.
Explore ATLAS · Qualified run and video
A constraint intelligence project that organizes operational bottlenecks, evidence gaps and validation paths. Public TypeScript source and project documentation describe the implementation.
A Python marine intelligence project bringing environmental observations, risk context and decisions in the field into one workflow. Its public repository documents the architecture.
A Python/SQLite research project that structures creator content into narratives, references and analytical records.
The public repositories document these projects. Current deployments, pipeline health, dataset scale and predictive accuracy for Economic X-Ray Vision, TideLine and YFI have not been revalidated for this portfolio.
Tools used in the inspected releases: Python, JavaScript, Node.js, JSON Schema, HTML/CSS, Git and Markdown. Each repository documents its specific stack and dependencies.
Disciplines: evidence modeling, validation, local data integrity, reproducible experiments, control systems, technical communication and privacy.
- ATLAS: a qualified controller tied to telemetry, parameter identity and reproducible race evidence.
- FIELDNOTE: a synthetic workflow exercising stale-write rejection, revision integrity and persistence across saves, reloads and service restarts.
- ATM research core: conditional studies kept distinct from account backtests, paper records from live records, and software checks from evidence of an economic edge.
- Start with the operational question and its constraints.
- State what would disprove a hypothesis and preserve contrary evidence.
- Publish the procedure, assumptions and limits alongside the result.
- Share the engineering while protecting the people, accounts and records behind it.
Coordinates: Terra Incognita