Overview About Thirdeye Data Inc.

Build and launched into production an end-to-end platform that picks up third-party provided images of electric poles to detect the quality of the image and decipher tags on the poles for further downstream consumptions in analytical systems.  Various AI models are applied to detect various image parameters like: Occlusion Blurriness Angle detection Tags Deciphering

Over the years, Client has accumulated hundreds of ML models for various levels and types of predictions, object detections, tags identification. It is developing an exhaustive MLOps platform for streamlining all activities, for data scientists, data engineers, and business stakeholders.  Data LakeHouse - Design, Architect, Creation Data Pipelines - Ingestion, Cleansing, Transformations, Correlations ML Flow - Model Development, Training, Deployment, Versioning

Built an anomaly detection platform to detect/predict anomalies in various server-side metrics like CPU, memory usage, traffic in / out, etc.   Completely open-source based, and runs on Spark Clusters. The platform has the following capabilities: Can detect anomalies on more than 10,000 metrics in the given datasets In-built Auto-AI functionality  On detection, raise alerts in dashboards and send notifications. Ability to drill down to identify root causes.

Enhanced existing keyword based content search engine on the HUDOC site to an AI & NLP based Semantic Search engine.  Leveraging NLP algorithms, performed Topic Modeling on legal documents hosted on the Azure cloud.  Associated provided concepts with NLP identified terms. Performed Summarization, Custom Entity Extraction, Key Phrases Extraction, Title Generation on a per document basis. Enabled User Feedback loop with rating mechanism and Add/Delete functionality of Concepts.

Developed from scratch a search engine for real time and domain specific text search to find out relevant information about the threats / vulnerabilities affecting an enterprise. Used Natural Language Processing (NLP) - Topic modelling using LDA algorithm. Leveraged Solr as the main search engine for indexing the data. Used Python Flask as the MicroServices framework. All code was containerized using Dockers and deployed on the Kubernetes platform.

Architected, designed & implemented the next generation CyberSecurity platform which tracks network data to analyze real time threats. The System as developed: Performs real time analytics using Artificial Intelligence technologies on petabytes of network data to continuously predict the next cyber threat. Provides a persona based 360-degree view of the security aspects of the overall network. Delivers a Deep Learning based Analytical and Cognitive engine.

For Phase 1, ThirdEye worked on the following items: Migrate all data sets, including ones in Network Drives to the new Data Lake based on Azure Blob Storage. Migrate data from SQL Servers to the Data Lake. Cleanse, prepare, enrich and transform data as needed for downstream consumption.  Implement data security structure and policies. Define & implement data governance policies.

Developed an Open-Loop-System that aids operational personnel to control the glass coating process, improves the product quality, and reduces waste.  The Open-Loop-System receives live data, computes the predicted end-of-line metrology values and corrective action suggestions for the process parameters in real-time, while panes are being coated.  The predicted metrology values and parameter suggestions were made available in real-time on the shop floor in the form of a graphical user interface, with which the operational personnel interacts. 

Developed an AI based knowledge management system that enables its users to identify relevant Subject Matter Experts (SME) on various topics of current interest by asking simple questions and getting detailed responses with a listing of ranked SMEs.  Developed deep JSON based profiles for every employee. All interactions were through a Chatbot interface. Users gave English language queries seeking for SMEs on various topics.

Extended NimbleStorage’s current Predictive Analytics Platform to predict in real time using 10x additional data points. Ingested 10x additional sensor data points into the system. Increased rate of data ingestion by more than 50%. Achieved near real-time data analytics capability. Improved analytical queries response times ability by more than 900%.

Setup data lake that ingests 200+ data sources with varied data sources types and formats. Setup data pipelines that ingests data in batch, real-time and incremental basis. Developing ML models to deliver inventory optimization strategies that will inherently help the company's bottom line. Will develop a Predictive Maintenance to be used by the Manufacturing Division  Will develop a Sales Forecasting system that the sales and marketing team would be using.

Developed Amazon AWS's first solution on the AWS Marketplace.  The solution showcased:  Integration with multiple services (both native and third-party ones) as available on the AWS marketplace. Used machine learning models for predictive analysis. Created analytical and predictive dashboards for visualizations. Developed mobile app to get notifications based on the predictive analysis. Customers deployed the solution with a single click and had it up and running within minutes.

Developed end-to-end data pipelines for operationalizing AI  Solutions.  Leveraged ML to deliver maximum value to patients, ensuring that medications are given to the right patient, at the right time & at the right dose.  Used Deep Learning techniques to accomplish the image recognition of medical pills through a Smart App. Used Sentiment Analysis for capturing patient’s experience, both online and using the connected devices.  Ingested real-time and batch data into a HIPAA compliant data lake.   Analyzed all data using D3.JS based analytical dashboards. 

Analyzed the enterprise datasets and external data sets that impacts pricing. Developing data pipelines for ingesting enterprise datasets and external data sets that impacts pricing. Developing Market Intelligence KPI Dashboard with deep analytical capabilities.   Will develop Machine Learning models using Enterprise Data sets. Will Augment Enterprise Datasets based ML models with External Datasets. Will Operationalize Pricing Recommendation Models across the entire enterprise.

Developed a world class Big Data analytical product, Navik Converter that helps enterprises to run successful marketing campaigns for specific business goals. One of the first products to be built on the Google Cloud Platform, the product leveraged state of the art Big Data technologies.  The Analytics algorithms were developed as MapReduce programs dovetailed together to enable business users to make business decisions with insights and decision metrics.

Electric poles carrying electricity for Energy Utility companies often develop snags that restrict the flow of electricity and causes wildfires resulting in losses for both the company and consumers. Developed a solution that would: Identify various images in a picture, including electric poles Detect anomalies in these electric poles and their sub-parts Predict the probability of defective electric poles to further deterioration  Images captured from drones flying over the electric poles were ingested into Azure blob for anomaly detection. The solution was used by Microsoft’s customers, partners and field sales personnel worldwide.

Developed an end-to-end analytical & predictive solution that enhances Customer Experience using Omni-Channel presence based on Microsoft Azure services. The solution connects the in-store and online experiences of the customers and highlights the difference between a Connected Consumer and a Traditional Shopper.  Provides dynamic recommendations and a 360 degree cyclic data exchange. Zero touch purchasing experiences with secure transaction authorization using Face Recognition.

Developed end-to-end Campaign Conversion Platform that optimizes campaigns for converting users to paid subscription plans using Machine Learning techniques.  Developed hyper targeted campaigns for individual users based on their profile that has been curated over time.  Improvised ML models by incorporating a complete feedback loop.