Six eras, one throughline.

This is the earliest, and in some ways the widest-ranging, stretch of the career: embedded software and algorithms for ADSL modems and digital audio broadcast (DAB) systems, computer-based automation for acquiring and processing images from scintillation cameras and sonogram devices, and a computer-based supervision system for the industrial process of a cement factory.

Three domains that share almost nothing on the surface — telecom signal processing, medical imaging, and industrial process control — but all required the same underlying skill: getting software to correctly and reliably drive real, unforgiving hardware.

Designed and implemented Fax-over-IP protocols for DSP-based telephony systems, along with the diagnostic, testing, and validation tools needed to prove protocol correctness and system stability. The optimization work pushed modem density on constrained hardware to five concurrent modems on a single DSP 56311.

Read the fuller story: Five modems, one DSP.

Developed core software for high-throughput fixed-wireless broadband systems, both Point-to-Point and Point-to-Multipoint, supporting data rates up to 200 Mbps. Designed and implemented cryptographic and security modules aligned with FIPS requirements, and architected key components — including the memory manager and packet classification system — for the high-throughput Point-to-Multipoint platform.

None of it could have started without building low-level tooling first: an assembler and debugger for the custom Point-to-Multipoint processors, since nothing like that existed for the hardware yet.

Read the fuller story: Building the tools before you can build the system.

First stretch at Bloomberg, working across a wide set of backend systems: querying and processing large operational datasets for the internal ticketing system, reporting and analytics pipelines for operational visibility, and a feedback and survey management system supporting performance evaluation of customer service representatives.

Two pieces of this era connect most directly to what came later: designing and executing the data migration and validation workflows that moved proprietary systems to SQL Server while preserving data integrity, and building the backend services for the fixed-income product description system (DES), covering government, corporate, and structured instruments — plus ETL pipelines and automation for index-rebalancing workflows that reduced manual effort and operational risk.

This is where the throughline of the later Bloomberg work — correctness and validation as a first-class concern, not an afterthought — starts to show up clearly.

Between the two Bloomberg chapters, an eighteen-month stretch running an independent consultancy — owning system design, implementation, and deployment directly for clients, across three unrelated domains: a weight-measurement system for an industrial packaging line, a scheduling and workflow system for diagnostic medical imaging clinics and mobile centers, and a telecom invoicing and data-processing platform handling large datasets and operational reporting for a telephone service provider.

The common thread across all three: translating an operational problem — one that existed as a manual process or a legacy system — into a dependable, working piece of software, without a larger engineering organization to lean on.

This wasn't the last time going independent would happen — see Logic Soft Design, 2026–present.

Eleven years back at Bloomberg, this time with ownership over architecture rather than individual backend systems. Two projects define this era:

Redesigning the Yield & Spread (YAS) analytics system — used globally for fixed-income analysis — to support portfolios of securities instead of single securities, which meant leading the decomposition of a monolithic system into distributed microservices: defining service boundaries and data contracts, then introducing parallelized computation and optimized data flows. Result: a 25% reduction in latency, and a system that could be tested and deployed piece by piece.

Architecting and delivering a redesign of the fixed-income benchmark calculation system — the project I'm proudest of, and one where the design was entirely my own — cutting processing latency by 85% while improving scalability and reliability, and building the validation and reconciliation framework that reduced post-deployment defects by 15% and kept the analytics consistent across every Bloomberg application that depends on them.

Later in this era, that also meant leveraging AI-assisted development tools to accelerate design iteration and improve test coverage — treating them as another instrument for correctness and speed, not a shortcut around either.

Read the fuller stories: the benchmark system and the YAS decomposition.

Left Bloomberg in February 2026 after eleven years owning the fixed-income analytics platform. Currently doing two things at once: building Logic Soft Design, an independent practice for software consulting and development, and evaluating the next senior systems engineering role.

This isn't the first time going independent — see LogicSoft Design, 2013–2014 — but it's the first time doing it with more than a decade of large-scale, high-stakes systems work behind it.

If you're working on a hard backend, distributed-systems, or financial-analytics problem — as a client or as an employer — get in touch.