Peter's AI Projects Demo

Interactive AB Testing App Using Probabilistic Programming and AWS

I build interactive machine learning APPs using python web framework Django, Docker, version control and cloud managed services

Here demo my owned AB testing APP. The algorithm I used is called probabilistic programming which compares the probability significance of the treatment sample against the controlled sample. This method is considered to deliver more reliable AB testing compared to the traditional null hypothesis.

The tech stack of the app is Django, plotly.js, Docker, CI/CD tool of GITHUB Action and AWS managed APP engine Beanstalk with autoscaling.

Supply Chain Route Optimization

One of the most common supply chain optimization tasks is vehicle routing, in which the goal is to find the best routes for a fleet of vehicles visiting a set of locations.

There are different approaches to address the routing optimization challenges. Some widely used algorithms are dynamic programming, mixed integer linear programming, et al. Here I demo a simple case using the dynamic programming to help a salesman to plan travel on 13 cities.

The demo here is an interactive visualization using folium package based on the solution of dynamic programming.

Price Setting, Optimization and Elasticity

Price setting refers to the process of determining the optimal price at which a product or service should be sold to customers. It involves analyzing various factors such as production costs, market demand, competition, and customer behavior to arrive at an effective pricing strategy. The goal of price setting is to maximize revenue and profitability while ensuring customer satisfaction and market competitiveness. Price optimization generally can lead to 3 to 5 percentage revenue increase.

I extract the elasticities of SKUs using demand modeling and machine learning. I also optimize the SKU prices using the latest Constraint Optimization algorithm such as Google OR tools.

I have successfully implemented pricing solutions in leading groceries, e-commerce, transportation broker and fast food chains.

Large Language Models

I fine tune and train the large language models for different business use cases

The technology I use is Huggingface Transformer, Tensorflow, and GPU infrastructure from GCP and AWS

Business Intelligence Dashboard Using Sigma Computing

Business intelligence is essential for informed decision-making, enabling organizations to analyze data and gain valuable insights, optimize operations, identify opportunities, and stay competitive in a data-driven world.

Here is a dashboard demo of auto claim data for major automakers. The BI tools used here is Sigma Computing

Risk Forecasting using Tensorflow Probability Monte Carlo

Risk forecasting is important for business operation of many industries such as supply chain, transportation, finance and clinical operations, et, al.

Risk forecasting is usually hard to solve as lacking information and future is unpredictable for most cases.

I implement the latest Tensorflow Probability Algorithms to conduct advanced Monte Carlo simulations based on probability programming. The algorithm is able to simulate random walks based on limited historical data that have no clear patterns.