Steven Ellis
Albany, NY
0→1 Product Builder | AI-Assisted Prototyping | Operational Workflows
I turn messy operational problems into working software.
My background started in military operations, moved through organizational leadership and government analysis, and eventually into software. The through-line is simple: understand how people actually work, find where information or responsibility breaks down, build something, put it to work, learn from it, and make it better.
Selected Product Work
AEGIS
AI-assisted operating environment for coordinating complex product work
Problem
Hypothesis
Prototype
Iteration
Lesson
Problem
Fragmented workflows and dispersed information create an incomplete operational picture. When research, engineering, and communications happen in silos, decision support reflects that fragmentation rather than organizational reality.
Initial Hypothesis
If we treat AI as an operating environment rather than just a chat window, we can unify product inputs into a single source of truth for decision support.
What I Built
Aegis brings together research, architecture, engineering workflows, QA, pressure testing, tasks, communications, project health, decision support, and executive briefing into one functioning interface.
What Changed
During development, I shifted from building one 'do-everything' AI assistant to a collection of specialized AI roles. Now, specialized models handle information extraction, risk synthesis, and brief generation separately.
Product Lesson
The power of the tool is not in the model's intelligence alone, but in the structure of the data it is allowed to see.
Hudson Valley Warriors HQ + Funding Ecosystem
Turning participant-level inputs into organizational action
Problem
When a member says 'I need a helmet,' it’s a logistics task. When fifty members say it, it becomes an equipment requirement. Without a system to unify those inputs, the organizational funding and operational decisions remain disconnected from the actual needs on the ground.
Prototype 1: The Input Loop
Member equipment request → Equipment Needs Queue. This automated the intake process, moving requests from member profiles into a structured operational environment.
Hypothesis
A unified data pipeline can transform individual requests into real-time organizational requirements and direct funding targets.
Prototype 2 & 3: Requirements to Funding
Member request → Equipment requirement → Organizational funding need → Funding gap → Funding platform.
Prototype 4: Closing the Loop
Saved funding opportunity → Deadline → Operational calendar. Grant deadlines and reporting requirements auto-populate the team's operational timeline, ensuring continuous funding readiness.
Tradeoff
I prioritized database structure over a custom front-end UI for the early prototypes. This allowed us to iterate on the operational workflow and organizational logic faster than if we had committed to a specific interface too early.
Product Lesson
Operational software is only as good as the least accurate piece of data people are willing to input.
Current Toolkit
AI
ChatGPT, OpenAI, AI-assisted development, agentic workflows
Build
Base44, Wix
Data & Operations
Airtable
Integration
APIs, webhooks, automation
Infrastructure
Cloudflare
I continue expanding the technical side of the toolkit as the products require it.
The Through-Line
Before I built software, I spent years inside operational systems. In the Army, that meant command posts, CPOF, readiness, assets, teams, training, and decisions made while the situation changed. In government, it means regulations, evidence, organizational boundaries, handoffs, and fragmented information.
That’s the work I want to keep doing.
Now I get to work on the other side of the problem: building the system people use to make sense of it.
Steven Ellis
Albany, NY
LinkedIn: https://www.linkedin.com/in/steven-ellis-908031b9/