Bodies of Work

Research doesn’t stop being research once it starts paying rent.

3 research programmes run alongside a production engineering practice — internal products and enterprise engagements built under real operational pressure. The publications below are where the thinking, in both directions, gets written down.

3 Research Programmes · 6 Professional Engagements · 12 Publications

Research Programs3 Programmes

Status: Research & Development
Duration: Since 2026

Transit Intelligence

It ingests live vehicle telemetry alongside the static schedule and asks a simple question continuously: does the vehicle currently agree with the plan? When it doesn't, the drift surfaces on a dispatcher's screen before a rider is left standing at a stop that already happened.

Operational Platform

Where dispatchers watch the gap between plan and reality widen in real time, and decide what to do about it.

Go Event Lab

Understanding event systems before building production systems — high-throughput ingestion proven against synthetic load first.

Lakehouse Engineering Lab

Makes months of positional history queryable in seconds, so a drift pattern is traceable, not just visible today.

Distributed SystemsData EngineeringOptimization

Connects to Stratum

Status: Active

Stratum

An edge analytics engine — local-first, running entirely on browser storage and WebAssembly to reproduce the base of what a tool like PowerBI does: analytics, mapping, a full data engine, with no server round-trip in the loop. A text-to-SQL layer and a WebMCP connection are planned above it, but those are additions to the engine, not the engine itself. Its rendering principles carry the visualization layer for Transit Intelligence, and a more constrained variant — scoped to company work only — powers the frontend of Quant and Track.

Edge ComputingLocal-First SystemsAnalytics Engines

Connects to Transit Intelligence · Quant · Track · Pulse

III

Status: Completed — ETH Zürich

Network Extinction Dynamics

Master's research at ETH Zürich — a real-world simulation of plant–insect networks across multiple scales, timelines, and levels of modularity, built on trait-matching and decay algorithms. The work measures how network structure changes under cascading extinction: which factors drive individual species loss, and where the tipping points sit — the thresholds past which a network can no longer recover.

Computational EcologyNetwork ScienceSimulation

Professional Engineering6 Engagements

Engineering

Quant

Active research

A backtesting framework that replays trading strategies as event streams, so a strategy's performance is judged by how it responds to sequence and timing — not just by its final number.

Quantitative SystemsConnects to Stratum

Pulse

Active

An analytics layer sitting on top of operational systems — CRM, ERP, and project management software (current integrations: Tally, Autodesk Construction Cloud). Data flows through a medallion architecture into a warehouse, landing in gold-layer tables scoped per business. The pipeline is owned and run server-side; dashboards render directly from the gold tables on the frontend, using the same rendering principles as Stratum.

Analytics InfrastructureConnects to Stratum

Nudge

In production

An event system that turns platform webhooks and scheduled batch syncs into role-gated alerts, decoupled from the request path that triggers them and tolerant of both real-time and delayed delivery windows.

Reliability Engineering

Track

A monitoring layer that turns raw field and activity data into forecasted progress, so gaps surface before a deadline does.

Operational AnalyticsConnects to Stratum

Consulting

Enterprise Construction Cloud Rollout

Co-led a full-scale rollout of a cloud construction management platform for an enterprise construction client, building 50+ reporting dashboards alongside the technical transition over a five-month engagement.

Enterprise Transformation

Enterprise Productivity Monitoring Suite

A multi-tiered dashboard suite spanning project, labor, and activity productivity for an enterprise client, with forecasting built in to surface risk before it becomes delay.

Business Intelligence

Client names and internal architecture are withheld by agreement — these are described by function, not implementation.

Engineering Library

12 Publications

EssaysResearch NotesArchitecture Notes

Enter Library →