Our Work

“Profitics is more than a technology company. They are a trusted partner that takes in all the complexities of the problem and responds with a technology solution. Once they are involved, the problem is as good as solved.”

Shae Hinson, President, Accordus

“We manage over 40,000 profiles and at peak stream 60 games simultaneously, We are very impressed with the cost effective reliability and scaling that Profitics’s solution delivered for us.”

CTO Maxxathlete, Mr. Del Riego

“We engaged Profitics in September and the first version of Profitics’s solution went live in December. We have come to rely on Profitics to provide game-changing solutions and rapid results.”

Tom Bartlett, President, CPRS

“Profitics delivered cutting edge computer vision expertise and ML models for various projects over the years with rapid prototypes at an impressive velocity. For cutting edge technology, Profitics has been a reliable partner.”

Venkat Yetrintala CTO Monotype

Brand Theft Mitigation

Technologies:
Anthropic API, Google API, OpenAI API, Python, Azure, CockroachDB, Airflow.

Solutions:
• Image based feature engineering effort was extensive to identify the aspects of an image that will differentiate between a morphed image of a fake brand and a worn out logo on a physical product.

Our customized anomaly detection AI network accurately identifies a fake product.

We built and tuned a subsystem to generate synthetic data for simulating possible genuine logo damage.

Retail Beverage Brand Mix

Technologies:
OpenAI API, AzureML, Google VisionAPI, AWS, CockroachDB, Airflow.

Solution:
• Extensive work in image pre-processing and creation of a Chroma vector DB for Brand and Brand related metadata.

• Object identification, brand identification, OCR price reading, package size recognition problems were solved to make this a successful solution.

• BevAnalytics saved more than 60% of their core operational cost in data collection.

Software Architecture Review

Technologies:

OpenAI API, AzureML, OLlama Local Install, AWS, MySQL, Python,AirFlow.

Solution:
• Made Architecture reviews and compliance self service adding velocity to the deployment of over 140000 applications.

• Architecture review documents and their inputs were used to create a Chroma vector DB of Architecture reviews.

• Used Llama Tiny as the base for fine tuning using Hugging face transformers.

Deep Bayesian Neural Network

Technologies:
R, Matlab, Kettle, C++, Java.
Solution:
• Built and tuned a deep network (4 hidden layers) fixed income instrument pricing, used Yahoo Finance Data set for training and tuning the network.

• Extensive experimentation with features (duration, convexity, vega etc.,) and network depth (all the way to 32 layers).

• Results of Monte Carlo simulations used in training multilayer perceptron (MLP) networks with backpropagation, compared with radial basis function (RBF) network.

Gale-Shapley Matching

Technologies:
R, AzureML, Azure, Azure Storage, C#, ASP.NET, BootStrap, Angular, JQuery, TFS, SQLServer, SSRS, SSAS.
Solution:
• Implemented Many to Many Matching Algorithm by extending Gale-Shapley in R.

• The platform is an extensible HIPAA compliant data collection platform for families that are interested in adoption.the platform has a workflow component that integrated various steps of adoption across many organizations.

• Setup a continuous delivery automation on Azure.

Quant Models for Trading

Technologies:
R, Matlab, Kettle, C++, Linux/Unix, C#, SqlServer, Cplex, kdb+, LIM, HDF, JQuantLib, Python.

Solution:
• Built numerous analytics models in Java, C++, Matlab, R for options, and derivatives.

• Macro alpha strategy based on currency market. (2011 – present). Macro alpha strategy based on ES/GC (2010 -2011) .

• Quant2Xchange: Developed a high speed interface between Matlab/R and Fix Engines (TT Fix Adapter) in Java.