Successfully Deployed Projects

Scroll through some projects that have been completed end-to-end and are currently in use.

AI-Driven Customer Reorder date Prediction

For one of asia’s leading healthcare service providers

This solution leverages partially agentic AI by autonomously retraining models, detecting data drift, and generating predictive insights, enabling proactive customer engagement with minimal human intervention. The continuous cycle of observation and planning needs no intervention, all that’s left to do is act on the suggestions by calling.

The Objective

Develop a machine learning model that could forecast reorder timelines at a client-product level, allowing proactive engagement with customers before they made purchasing decisions elsewhere.

The Challenges

  1. Data Complexity: Handling large-scale pharmaceutical sales data (200,000+ rows)

  2. Model Selection & Complete Automation: Ensuring the best-performing model is dynamically chosen based on new data.

  3. Scalability & Maintenance: Minimizing human intervention in model retraining and prediction workflows.

Architecture

Connecting to the existing database and effectively transfer this massive data for the next steps while being mindful of resources.

Use preprocessing methods to maintain and create useful features while removing unnecessary noise.

Also, separate the data into different sets for a better ML model output.

Each model trains on a specific dataset and learns the purchasing patterns respectively for each client and their specific order to predict when they will reorder.

This is done completely automatically.

Create a Docker container with the preprocessing, models and other failsafe methods like Input drift detection

Schedule 2 runs to run automatically:

  1. Retraining models as a new container image - every 6 months

  2. Serving Predictions on new data - every week

Hosting the entire framework on Google Cloud Platform using Vertex AI

Business Benefits

Currently in production and being marketed.

Well timed telemarketing calls will ensure client retention.

Deviation from the predicted reorder date will ensure follow ups to address why a client has not reordered.

Other Projects

Dashboard for Leadership Et` Humanite`

Scale based Survey analytics dashboard tool that dynamically works on new csv files. Connected to Google Drive.

NLP Code Converter for a prominent data analytics company

An internal tool to help employees convert the well known Regex code to Spacy, which is sometimes used on entity tagging software.

Agentic AI solution for system file retrieval

An Advanced Windows File Search System with Intelligent Decision-Making and a full fledged workflow for document enchancements

SDLC for hotel cancellation prediction

This project leverages machine learning to predict hotel booking cancellations, helping hotels optimize revenue and resource management. Built using Agile and Scrum methodologies.

Intelligent AI Research & Analysis Agent

An AI-powered data extraction and intelligence agent that connects to internal/external systems to automate insights, streamline decision-making, and enhance operational efficiency.

Demand Forecasting for Freight Allocation

A forecasting engine to predict future freight load volumes based on historical data. This ensures optimized truck allocation, reducing inefficiencies and improving logistics planning.

Custom File Retrieval Automation for a manufacturing company

A custom automation tool that streamlines file retrieval based on dynamic data criteria, enhancing efficiency and decision making.