Project Overview
This case study focuses on my approach to systems architecture and technical program management: translating requirements into workstreams, defining interfaces and restructuring development dependencies to protect a milestone.
Capabilities Demonstrated
[1]
Translating system requirements into workstreams with defined responsibilities and deliverables.
[2]
Restructuring development to enable parallel work and resolve blocking schedule dependencies.
[3]
Using simulation and subsystem testing to verify functionality before full-system integration.
[4]
Defining system architecture through component boundaries, data flows and technical interfaces.
PROJECT CONTEXT
System Purpose
Materials Procurement and Manufacturing Planning
By integrating price forecasts into existing procurement workflows, teams are better able to evaluate when to purchase materials, how much inventory to source and how price changes could affect project costs. The system is most useful for long-lead procurement where purchasing decisions are made months before materials are needed.
Consulting and Financial Analysis
How It Works
The system works by identifying basic market components, and grouping certain relationships into “reaction fingerprints,” i.e. recurring patterns in how prices fluctuate over time. Then, a combination of statistical analysis, deterministic analysis and machine learning inference evaluates these reactions in context to generate and combine signals. Finally, the outputted signals inform likely direction and potential extent of price movements within certain confidence boundaries.
Personal Scope
I co-founded the company, hired a small team and raised seed funding for the development phase between June 2025 and December 2025. During development, I led end-to-end program management including concept of operations, requirements definition, system architecture, workstream decomposition, testing and milestones across the project’s R&D phase spanning January 2026 to September 2026.
PART I
System Architecture & Workstream Definition
Table 1.1: Workstream Responsibilities and Deliverables
Ownership & Collaboration
Colors indicate my ownership and collaboration within each workstream.
Owned by me
Led by me with business direct report
Collaborated with ML engineer
Owned by ML engineer
1.2 System Architecture
DIAGRAM KEY
Workstream component
External source or platform
System boundary
Information exchanged
Figure 1.1: System Architecture and Workstream Interfaces
The diagram above maps the technical workstreams to the system’s functional components, with workstream IDs linking each component to the table above.
I defined the interfaces around the information passed between components, including feature definitions, training examples, trained models and structured features.
The arrows show these handoffs and distinguish the model-training path from the operational path, where structured features feed the decision engine alongside the trained model and defined rules.
Program management (WS1), systems architecture (WS2) and verification and validation (WS8) span the system and are therefore not shown as individual processing components.
PART II
Program Management Case Study
This section highlights one decision from the broader program to illustrate my approach to dependency management and development planning within program management workflows.
Restructuring Dependencies to Protect a Milestone
2.1 Surfacing a Dependency Problem
The original development schedule included some float to absorb delays while meeting the proof-of-concept milestone. Delays in high-speed processing development were consuming that float and leaving insufficient time for downstream testing. Testing of the decision-making layer depended on high-speed processing readiness, so further delays would push that work back and threaten the milestone. The development sequence was workable, but the remaining schedule required an approach that allowed both efforts to progress in parallel.
2.2 Resolving the Dependency Chain
I reviewed which aspects of decision-layer testing required high-speed processing and which could proceed using historical and delayed live data. I introduced software-in-the-loop (SIL) simulation to test the decision-making layer while high-speed processing development continued. This removed high-speed processing readiness as a prerequisite for that testing and allowed the two efforts to converge later for full-system integration.
The side-by-side diagram shows how this change replaced the original sequential dependency with parallel development paths.
Original Dev Sequence
Sequential development
Revised Dev Sequence
Parallel development with shared integration
Figure 2.1: Development Dependencies Before and After SIL
Both sequences begin with an initial decision framework available for testing, combining signal and rule development (WS3) with ML development and integration (WS6). The comparison shows how SIL allowed decision-layer testing under WS8 to proceed alongside high-speed processing development (WS6). Both sequences retain full-system integration and testing (WS8) as the final milestone.
2.3 Implementing SIL Approach
The SIL approach supplied historical and delayed live data to exercise the decision-making layer before the high-speed infrastructure was ready. I used subsystem-specific acceptance tests in Python to evaluate decision accuracy while the infrastructure work continued against its latency requirements. This allowed the decision-making layer to be tested and refined ahead of full-system integration. Full-system integration testing remained necessary to evaluate the combined system’s behavior and performance under live operating conditions.
2.4 Outcome
Using SIL simulation allowed decision-layer testing to progress while the real-time infrastructure was still in development. This preserved time for refinement before full-system integration and helped keep the proof-of-concept milestone achievable. The completed system was subsequently validated through live operational testing and delivered to stakeholders.
Anderson Todd
Engineering Portfolio