This is a portfolio summary of a work project — proprietary internals are intentionally omitted. ← Back to portfolio
Data Analysis Agent

How it's structured

An agent that processes time-series test data to identify points of failure, cross-references the corresponding image captured at each point, and writes a report flagging anomalies and low-confidence results — built during my time as a hardware/test engineer.

Note on Proprietary Information:
Developed for enterprise medical device testing. System details and architecture are presented at a high level to protect proprietary data schemas, internal tools, and company IP.

System Architecture

Flags a failure point in the data, then pulls up the image that goes with it.

Test run data
Time-series sensor logs + captured images, timestamp-aligned
CSV / log files
↓
Signal analysis
Scans each channel for failure points and low-confidence regions
PythonPandas
Image cross-reference
Pulls the corresponding image at each flagged timestamp as supporting evidence
↓
Reasoning agent
Reviews flagged points + images together and drafts findings in plain language
LLM Agent
↓
Confluence report
Objective, methodology, results, and recommendations, with links to raw signal data and image sub-sheets
Confluence API
JMP workbook
Summary metrics, interactive graphs, and formulas compiled for team review
JMP

Software Used

Python Pandas NumPy LLM Agents Data Analysis Confluence API JMP

Illustrative Example

Synthetic data, shaped like a real report — not an actual test run.

Channel A — Pressure
t = 04:12.8
Flagged — high confidence
Channel C — Temperature
t = 11:47.3
Flagged — low confidence
Channel B — Flow Rate
t = 18:02.1
Flagged — high confidence