The opinions expressed by Entrepreneur contributors are their own.
You are reading Entrepreneur India, an international franchise of Entrepreneur Media.
“The convergence of machine learning and real-world industrial applications requires solutions that go beyond theoretical frameworks.” This echoes the opinion of Tharakesavulu Vangalapat, a senior principal data scientist whose patent portfolio spans predictive maintenance, intelligent lighting systems, and enterprise analytics platforms. “The seven granted patents represent years of success in translating complex algorithmic concepts into concrete systems that organizations deploy at scale.”
His work comes at a pivotal moment for the commercialization of artificial intelligence. According to the World Intellectual Property Organization, the number of global AI patent applications will surge to 89,000 in 2024, a 42% increase from 2020 numbers.
The predictive analytics market alone is expected to reach a valuation of USD 12.4 billion by 2024 and grow to USD 38.1 billion by 2030. Vangarapat's contribution sits squarely within this transformation, bridging the gap between academic research and implemented industrial systems that produce measurable economic benefits.
Vangalapat's career trajectory illustrates the evolution of applied AI over the past two decades. The transition to Signify (formerly Philips Lighting) North America Corporation marked a decisive shift towards invention. From 2016 to 2021, Vangalapat led the development of the Interact LightPlay application, an AI-powered platform that transforms static lighting into dynamic, responsive environments.
The system utilizes computer vision inputs and machine learning algorithms to enable real-time personalization and has achieved global recognition with over 1 trillion interaction metrics. Patents filed during this period addressed interactive color selection using dynamically color-changing LEDs to solve the technical challenge of translating user gestures into precise lighting adjustments across a distributed network. His predictive maintenance model for connected lighting systems has reduced downtime by approximately 25%, leading to significant cost savings for commercial deployments.
Financial services analysis and regulatory compliance
Mr. Vangalapat joined Broadridge Financial Solutions in June 2021 with responsibility for enterprise AI strategy across capital markets infrastructure. The company processes more than $10 trillion in fixed income and equity transactions annually and serves more than 1,200 financial institutions. His work focused on implementing machine learning systems capable of handling regulatory compliance, proxy voting analysis, and asset management forecasting on an unprecedented scale.
The World Demand Forecasting Model represents his most important contribution to financial analysis. The platform leverages ensemble learning techniques and time series algorithms to predict assets under management and net flows with a level of accuracy that enables US$4 million in annual revenue growth. A published methodology paper details a technical approach that demonstrates how Bayesian optimization and feature engineering address the non-stationary characteristics of financial data.
Long-term forecasts target US$60 million in strategic growth and position this model as a cornerstone of Broadridge's analytics portfolio. Financial services face increasing regulatory complexity. The Securities and Exchange Commission processed 7.2 million EDGAR applications in 2024, up from 5.8 million in 2020. Vangalapat built an intelligent document processing framework for DEF 14A and 10-K forms to automate the extraction of more than 100 data points per document.
Natural language processing and generative AI techniques have saved thousands of hours of manual processing time and approximately $500,000 annually, while reducing error rates by more than 90%.
His Customer Policy Vote Prediction Engine combines machine learning, natural language processing, and generative AI to automate shareholder vote analysis across proxies. The platform serves over 200 institutional customers, improving the accuracy of investor decision modeling, and has cumulative customer impact of over $100 million. This innovation establishes a new industry benchmark for regulatory compliance automation and reduces the turnaround time that previously constrained institutional investment strategies.
Research impact and cross-domain innovation
Vangalapat holds seven patents spanning multiple technology areas. The 2020 application covers the use of sensors and lighting to monitor egg quality and the application of computer vision to poultry management. The system identifies individual chickens based on biological characteristics, correlates egg data with egg-laying behavior patterns, and determines quality levels through an iterative machine learning process.
Eight citations from independent researchers demonstrate the adoption of this methodology across agricultural technology applications. Another family of patents focuses on coded optical communications for access control within secure environments. The lighting unit transmits a unique code via modulated lighting, allowing authentication by a light-receiving device carried by authorized personnel.
Plant health monitoring is a third area of innovation. Vangalapat's system synchronizes image capture with actuators that control irrigation, fertilization and lighting conditions. The processor analyzes visual plant quality across time series data and dynamically adjusts parameters to optimize growth results.
This framework gained wide recognition in controlled environment agriculture, where precise resource allocation directly impacts yield and profitability. Cannabis grading systems extend plant monitoring approaches with strain-specific quality assessments. Machine learning classifiers evaluate visual features against cultivation parameters and generate automatic grade assignments that help with pricing and inventory decisions.
Vangalapat's research has been published in IEEE journals and conference proceedings, and he has received 16 citations from independent researchers across computer vision, predictive maintenance, and agricultural technology. His Google Scholar profile records collaborations with researchers at MIT CSAIL, Signify Research, and Broadridge's analytics department. Peer review responsibilities include evaluating over 40 manuscripts for international conferences on AI and machine learning.
The 2025 IEEE 7th International Conference on Computing, Communication, and Automation recognized his contributions as a conference reviewer and his role in maintaining publication standards across emerging research fields.
Enterprise infrastructure and future direction
Vangalapat's technology leadership extends beyond individual algorithms to infrastructure, enabling sustained AI adoption. At Broadridge, we designed an MLOps pipeline that integrates continuous integration and continuous deployment workflows with automated monitoring and rollback capabilities. This infrastructure reduced implementation cycle time by approximately 40% while maintaining the audit trail required for financial services compliance.
Generative AI infrastructure presented distinct challenges. Deploying OpenAI GPT, Anthropic Claude, AWS Bedrock, and open source Llama models required guardrails to enforce constraints on personally identifiable information processing and regulatory compliance. Vangalapat designed an agent AI workflow that enables natural language interaction with the corporate knowledge base while incorporating safety mechanisms to prevent disclosure of sensitive data. Although the language model shows great functionality, it lacks essential understanding of an organization's data policies.
Guardrail systems must operate on multiple levels simultaneously, from input sanitization and output filtering to semantic analysis, to detect potential policy violations in the generated content.
Vangalapat's current research interests focus on multimodal AI systems that simultaneously process text, images, and structured data. Enterprise applications increasingly require integrated models that can handle diverse input formats without the need for separate specialized pipelines. Although underlying models pre-trained on large-scale, multi-format datasets promise generalization capabilities beyond domain-specific architectures, implementation challenges such as latency, cost, and accuracy thresholds remain significant.
Agentic AI frameworks represent another frontier. Autonomous systems that can plan, use tools, and iteratively improve enable complex workflows that previously required human oversight at every decision point.
