Insights
Niobia AI Blog
Perspectives on AI for advanced manufacturing, battery R&D, enterprise deployment, and technical workflows.
Will My Data Be Used to Train AI? A Practical Guide to AI Privacy Tiers
A practical guide to AI privacy tiers for teams using ChatGPT, Claude, Gemini, APIs, and cloud-hosted AI. Learn the tradeoffs between consumer AI apps, API access, and enterprise deployments for sensitive data.
Section
Battery Manufacturing
Electrode coating, calendering, and AI vision inspection for lithium-ion gigafactories.
AI in Battery Manufacturing Was Never a Switch
Most of what gets called AI in battery manufacturing is decade-old machine learning. What actually changed is the reasoning layer. The real architecture is five pillars with a brain that reads the floor and ledger, reads and writes the data and models, and never bypasses safety-critical control.
Chinese Gigafactories Run Under 10% Scrap. Western Plants Are at 30-40%. The Gap Is Not What You Think.
The yield gap between Chinese and Western battery gigafactories is not about cheaper labor or looser standards. It is about how process, materials, and quality data get connected and acted on. Here is what AI in a gigafactory actually looks like, the four levels of maturity, and where Western plants get it wrong.
AI Vision for Electrode Coating: 100% Inspection Cuts Scrap
100% web inspection vs manual sampling: 20x more defects caught. How detection accuracy translates to yield, and what coating defect patterns reveal.
Dry Electrode Coating: 8 Defect Modes and Their Signatures
Tesla's 4680 ramp exposed eight defect classes unique to solvent-free dry electrode lines. Where each fails, and why the visible defect rarely is the cause.
Calendering Defects in Li-Ion Electrodes: A Detection Guide
The six recurring calendering defect families, the process parameters that drive them, and how AI vision catches drift before scrap.
Section
Injection Molding
Scientific molding, process parameter interactions, and root-cause intelligence on the press.
What Is Injection Molding? An Engineer's Six-Phase Guide
97% of part quality variation is born in fill and pack. The six-phase cycle, the three pressures engineers conflate, and the real economics vs CNC.
The 12 Injection Molding Parameters That Actually Matter
Fill rate, melt temperature, pack pressure, gate seal — the 12 coupled parameters, what each controls, and why single-variable tuning fails.
Single-Variable Thinking in Injection Molding: Why It Fails
Injection molding parameters are coupled: tuning one shifts five others. Why univariate SPC misses drift and how multivariate signatures catch it.
10 Injection Molding Defects, Their Root Causes, and Fixes
Sink marks, warpage, flash, short shots, weld lines — each has a specific mechanism and diagnostic sequence. Most are misdiagnosed on the first attempt.
Live RCA Demo: Injection Molding Boss-Sink Diagnosis with AI
8D walkthrough of a recurring PA66 boss-sink defect, structured for IATF 16949 audits. What AI-assisted RCA actually changes and what it cannot replace.
Section
Design of Experiments
DOE methodology for advanced manufacturing: design choice, factor selection, and AI-assisted experimentation.
Section
AI & Enterprise
AI privacy, enterprise deployment, and data governance for technical workflows.