Research

I co-built an end-to-end Android, radio, and embedded-control prototype that translated real-time strategy interactions into drone commands.

The 2014 work established the command interface and execution path; Drone-Megaddon 2026 extends that foundation with a proposed language-and-vision architecture for operator-supervised coordination.

I coauthored a proposed extension of our built 2014 drone-control prototype: translate operator language into checked commands, ground references in video and track targets, then coordinate assignments across drones.

The architecture places language and vision models inside an operator-supervised system, with explicit choices about edge versus cloud execution, latency, bandwidth, and degraded connectivity.

An illustrative search-and-rescue trace shows how the layers work together; the paper reports no quantitative results from a deployed 2026 system.

I coauthored a framework for leading engineering organizations when AI makes code abundant but verification and judgment remain scarce.

It connects review theater, engineer development, and team incentives to practical changes: smaller reviewable units, scrutiny matched to consequence, evidence-backed gates, and outcome-based metrics.

An exploratory, correlational analysis of several thousand commits from a single organization illustrates why raw activity can overstate progress; the paper turns those observations into testable propositions about verification and apprenticeship.

I coauthored a decision framework connecting AI strategy to investment and execution: assess organizational readiness, trace business value through ROIC, phase investment through evidence-based gates, and synthesize a decision scorecard.

An eight-category risk assessment makes inference economics, vendor dependency, and organizational friction explicit.

Eight illustrative public-company cases, from Salesforce and Spotify to JPMorgan and NatWest, show how the framework changes investment conditions across deployment types; this is illustrative validation, not empirical validation.

The work was a Wharton independent study sponsored by Professor Gad Allon.