Projects

We love how intelligent technologies today can be applied to any industry. Here are overviews of some of our projects:
Cloud, Computer Vision
Accuracy of APIs

Among the various computer vision APIs built for cloud, it was challenging for our customers to determine which API’s they should use:

“Amazon’s? Microsoft’s? Others? Perhaps we should just stick with computer vision models...”

So, we conducted a scientific study of different vendor technologies. This included creating a custom dataset with ground truth annotation, defining metrics, building comparison tools, and analyzing results. We evaluated all aspects of each API’s functionality, including face recognition, emotion detection, gender prediction, and age estimation.

 

Ultimately, we found that different vendors excelled in different areas. We published our results in two conferences, and presented our findings to Microsoft, Google, and Amazon.

Automotive Industry, Manufacturing, Computer Vision
Mechanical Part Inspection

A tier 1 supplier to automotive companies, such as Toyota and Ford, faced challenges in manufacturing high-quality hardware fasteners. Since their industrial machine made over hundred fasteners per minute, manual inspection was impractical. They tried an automatic inspection system, but it was bulky and performed poorly.

 

The solution was built by our development partner, Trendzlink, using inexpensive, off-the-shelf cameras and processors. It runs a proprietary computer vision algorithm that scans a part in 100 milliseconds; hence, each fastener can be accurately scanned. The algorithm is easy to modify and can be applied to detect faults in other manufacturing parts as well.

Good
Bad
Automotive Industry, Manufacturing, Machine Learning
Warranty Data Analysis

Manufacturing companies spend billions of dollars each year to honor their warranty commitment and repair defects in their products. Since the type and quantity of defects are often incorrectly gauged, company service centers are challenged to stock up the right type and quantity of each part. This leads to repair delays, customer dissatisfaction, and can weaken quarterly financial results.

 

We developed an analytics solution to address this problem for the automotive industry. With mathematical models and machine learning, our solution can predict defects and costs with automotive warranty data. Additionally, the solution includes root cause analytics capabilities, which helps fix the problem and prevent future defects.

Health, Data Engineering, Natural  Language Processing
Group Therapy App

A health startup needed help for creating a mobile app for group therapy. The purpose of the app was to allow users to join communities they identify with and share their experiences without fear. This would complement existing mental health infrastructure by providing better accessibility and lower cost.

 

Our development partner, Coditation Systems, did the mobile app development, cloud data engineering, text classification, and sentiment analysis with deep learning models.

 

The solution exceeded expectations; in addition to meeting the startup’s needs, the app could also classify users based on the urgency with which they need care, all by discerning their mood with high accuracy.

Need more details? Contact us

© 2019 by Pifocal