Professional portfolio
I am James Dye, a Columbus-based AI-first systems builder and founder of XyFlow Innovations. My background combines GIS, survey field operations, UAS/LiDAR work, Python data pipelines, and agent-orchestrated software development.
XPRIZE milestone award
$36k
Solo Global Water Scarcity / desalination concept advanced to milestone funding.
Field operations
SAM
Crew Chief / UAS Operator working survey, drone, LiDAR, and utility environments.
Credential
Part 107
FAA Remote Pilot Certificate plus B.S. GIS from The Ohio State University.
Public proof
GitHub
Data pipelines, research systems, mobile apps, Unity tools, and product surfaces.
Selected artifacts
This page is for job applications and professional review, so projects are separated by signal instead of dumped into one mixed grid.
Public portfolio repo
Python ETL across GNSS clock products and space-weather feeds, with partitioned Parquet outputs and anomaly detection.
Runnable fixture suite validates parsing, episode detection, canonical Arrow schema, Hive partitions, and context enrichment.
Public curated surface
An AI-orchestrated research architecture built to stress-test agent collaboration, validation discipline, and current system limits.
Public repo ships the Control Room surface, doctrine, telemetry ledger, campaign reports, and evidence for how the agent workflow was managed.
Internal workflow tool
Lightweight field-crew automation for folder creation, photo placement, and naming conventions in survey workflows.
Built around real field friction: repetitive admin work, photo handling, crew handoff consistency, and office/field process gaps.
Private active build
A local pipeline for turning AI-assisted asset ideas into Unity-ready source, generated files, reports, and reviewable outputs.
Current workspace includes generated meshes, PBR textures, lab screenshots, variant runs, validation reports, and Unity import surfaces.
Public repo
Offline bird-sound trainer with real recordings, regional quiz packs, local progress, and source/license attribution.
Android release profile covers 1,084 species, 1,114 clips, local progress, quiz/study modes, and commercial-source audit tracking.
Public beta repo
Voice-first desktop app for long-form fiction and research with file-led storage, persistent project memory, and visible AI collaboration.
Ships Novel and Research modes, local filesystem project state, source-aware nonfiction workflows, and portable Windows packaging.
Field operations
The field background matters because many useful AI systems live where forms, photos, procedures, equipment, crews, and office systems meet.
Built a lightweight internal workflow tool to automate repetitive folder creation, photo placement, and naming conventions for survey field work. The value is straightforward: remove repetitive admin friction, make crew handoffs cleaner, and reduce the chance that field data arrives in the wrong shape.
External validation
Submitted a solo theoretical desalination design that advanced in the XPRIZE Global Water Scarcity competition and earned approximately $36,000 in milestone funding. The venture did not become a commercial prototype; the honest takeaway is domain acquisition, systems design, and competitive technical concept development under an external rubric.
Strong-fit roles
The core story is simple: I have already been building the kind of AI-native systems companies are now trying to figure out.
Read the methodPortfolio contact
For roles, collaborations, PackSmith support, Unity asset questions, or practical AI workflow work, email directly.