Teqniksoft is an international software development company with operations in North America, South America, and Europe. They provide product lifecycle solutions for desktop, Web 2.0, Cloud, and mobile applications, and core competency in embedded systems engineering. They provide consulting services for all stages of development from architecture through QA, including staff augmentation services. Their team has architected, designed, developed and delivered commercial products for large, public technology companies to early stage start-ups.
Teqniksoft’s Artificial Intelligence (AI) teams are composed of data analysts (DA) that combine deep theoretical knowledge in applied mathematics, statistical pattern recognition, and AI techniques together with hands-on, practical experience.
The DA teams are proficient in statistical tools like R and Panda (a Python module) and with Machine Learning (ML) tools like TensorFlow and Keras, as well as inference engines and fuzzy logic tools.
Teqniksoft’s teams have deep expertise in using AI to generate results and solve real life problems in multiple domains, such as (but not limited to) manufacturing environment, documents analysis, and parsing internet bots information.
For example:
- Fortune 100 company that produces precise devices that are assembled in a clean environment to prevent contamination. Teqniksoft DA developed models base on Fast RCNN and Mask RCNN using TensorFlow and OpenCV to identify if contamination exists, and what type of contamination it is with 98% accuracy.
- LawCarta.com provides an access to all supreme court cases for authors and professors that use this information for authoring books, writing course material, and research. These public information court cases are provided as raw texts. For smart search and categorization of these case laws, Teqniksoft’s DA developed a set of algorithms to clean the data, identify the justices, categorize the context and justice’s opinions for a better and more relevant search. Software used in this project includes: regular expressions tools, Clips inference engine, and TensorFlow.
- Reduction of quality assurance time by predicting for large manufacturers (50 million plus products per year) if product with its 4000 attributes will pass specific time-consuming quality assurance tests. Using R and Random Forest classification, the DA team achieved 91% accuracy in prediction causing 75% of product inventory to skip the test, while still ensuring 99% quality by augmenting information from partially overlapping tests.