IE PROFESSOR @ CPP | PRODUCT & SYSTEMS ARCHITECT
From ambiguity to working systems. Architectural clarity before costly execution begins.
My best work begins when something valuable needs to exist, but its purpose, requirements, or architecture are not yet clear.
I determine what should be built, align the people and disciplines around a coherent architecture, and lead the work from ambiguity to a viable result.
I apply this approach across physical products, software platforms, operational systems, services, and research-based technologies.
Since 2015
I designed a mixed-integer optimization architecture for scheduling a fleet of about 1,500 assets, using decomposition strategies and large-scale computational experiments to obtain exact solutions in under seven minutes. The work enabled near real-time re-optimization aimed at reducing downtime and improving fleet availability in asset-intensive operations.
2024
A failing prototype effort was turned into three working injection device prototypes by defining clear requirements, enforcing architecture decisions, and managing development through scoped execution and testing. The shift replaced trial-and-error with a controlled, validation-driven process.
Since 2025
RecomAid fixes how recommendation letters and qualification evidence actually get produced. It puts AI inside a structured, human-reviewed workflow, so professors and evaluators spend far less time, students get a clear way to prove their qualifications, and every decision stays consistent and accountable.
Since 2026
Alongwise turns career coaching into a working product: evaluate fit before you apply, generate tailored materials from your own evidence, and track every opportunity. Built with JSON-driven access control, Stripe entitlements, and a settings-backed LLM pipeline.
A shadow can be physically accurate and still lead us to infer a creature that never existed. What does this tell us about probability distributions, optimization models, simulations, and the limits of the frameworks we use?
LLM access alone does not create organizational value. The current phase of AI adoption focuses on Specific AI Services (SAIS): tools built around defined workflows, human judgment, measurable outcomes, and transparent costs.
Models are simplified perspectives on reality, much like shadows. They become dangerous when their limitations distort our understanding or begin to reshape the systems they were meant to describe.
What's been published or said most often is not necessarily what's accurate or complete. Mistaking "most common" for "true" is becoming a real problem.
AI does not help everyone equally. It disproportionately amplifies people who already learn by pattern recognition, curiosity, and first principles.
This paper is concerned with stochastic scheduling of production and maintenance activities in a permutation flow shop setting. We present a two-stage stochasti…
Construction projects usually get delayed for several time periods. When the planning horizon of a project is extended, projections for purchase and salvage of …
In this paper, we extend the formulation and solution algorithm of a Flight and Maintenance Planning (FMP) problem to cover a wider range of applications. We co…