eChapters

Human communication rarely depends on a single information modality. People simultaneously interpret language, images, sounds, gestures and contextual signals to understand their surroundings. Multimodal Large Language Models (MLLMs) seek to reproduce this capability by integrating text, vision and speech within unified computational frameworks. This chapter introduces the evolution and technological foundations of multimodal intelligence and explains how different data modalities can be aligned through shared representations. It examines the architectural components, training paradigms and evaluation methods used to develop multimodal systems. Applications are discussed across healthcare, autonomous systems, education, accessibility, conversational interfaces and intelligent assistants. For example, healthcare applications can combine electronic health records, spoken clinical interactions and medical images to provide richer contextual understanding. Similarly, autonomous vehicles can integrate visual road information, spoken instructions and textual navigation data. The chapter also considers challenges involving multimodal alignment, computational requirements, data quality, hallucination, interpretability and safety. It concludes by highlighting future research toward more natural, context-aware and human-centered AI systems capable of seeing, listening, reasoning and communicating

Large Language Models have moved beyond experimental research environments and now support conversational agents, coding assistants, recommendation systems and intelligent decision-making applications. However, delivering these services to millions of users requires extremely low latency, high throughput, scalability, security and cost efficiency. This chapter examines the engineering principles required for real-time LLM inference and large-scale deployment. It decomposes end-to-end latency into model execution, network and queuing components and discusses techniques for optimizing each stage. Model-level approaches such as quantization, knowledge distillation, LoRA, QLoRA and speculative decoding are examined for reducing computational requirements while maintaining model quality. At the infrastructure level, tensor and sequence parallelism, GPU and TPU acceleration, efficient batching, autoscaling and cloud-native deployment strategies are considered. Key-value caching and speculative decoding are discussed as important methods for reducing response time in interactive applications. The chapter also addresses privacy, confidential computing, real-time PII protection, energy consumption and monitoring of model performance. Practical deployment examples demonstrate how these techniques can support responsive, secure and efficient AI services at production scale.

The rapid growth of Large Language Models has demonstrated that increasing model size, training data and computational resources can significantly improve performance. This chapter examines the theoretical and empirical foundations of scaling laws and their relationship with optimization strategies used to train and deploy increasingly capable LLMs. It discusses the major scaling dimensions of model parameters, dataset size and computational budget and explains how predictable relationships between these factors influence model development. The chapter considers the foundational work on scaling laws and the subsequent movement toward compute-optimal training approaches. It also examines distributed training, advanced optimization algorithms, memory optimization, data curation, quantization, pruning and efficient inference. Particular attention is given to Mixture-of-Experts architectures, efficient attention mechanisms and inference-aware scaling. The interaction between scaling laws and optimization is presented as a synergistic process in which scaling principles guide resource allocation while optimization techniques improve computational efficiency. The chapter concludes by discussing emerging capabilities, sustainability, hardware requirements and responsible scaling of future foundation models.

There are many types of intellectual property and its type varies from country to country. The most common types are patents, copyrights, trademarks, design and trade secrets. Intellectual property is an intangible creation of the human intellect and this intangible nature of intellectual property faces problems in comparison of traditional property like land or goods. Unlike traditional property, intellectual property is not dividable therefore a number of people can consume an intellectual good without its consumption. In addition, investments in intellectual goods suffer from appropriation problems, for instance Landowners can surround their land with a robust fence and hire armed guards to protect it, but creators of intellectual property cannot do that to stop their first buyer from replicating it and selling it at a lower price.

Biological Diversity Act, 2002 was passed by Indian government in pursuance of Convention on Biological Diversity, 1993 The Convention on Biological Diversity (CBD) is a legally binding multilateral environmental agreement that has 194 contracting Parties (Countries) as its members with three objectives mentioned below:

1. Conservation of biological diversity
2. Sustainable use of the diversity and
3. Ensuring fair and equitable sharing of benefits of such use.

The recent concept of intellectual property emerged in England in the 17^th & 18^th centuries. The term Intellectual Property came in use in the 19^th century and in the late 20^th century. As of now the Intellectual Property has occupied common position in the global legal systems. There are different forms of Intellectual property rights which include patents, copyright, industrial design, trademarks, plant variety rights, trade dress, geographical indications and in some jurisdictions trade secrets too

The ITPGRFA is a comprehensive international agreement which is poised at guaranteeing food security through the conservation, exchange and sustainable use of the world's plant genetic resources for food and agriculture, instead of fair and equitable benefit sharing arising from its use. It also advocates Farmers' Rights.

13.1. Genesis of the Treaty

The conservation and sustainable use of plant genetic resources for food and agriculture are keys to feed global growing population in the coming future. The Commission on Genetic Resources for Food and Agriculture was established in 1983, and the voluntary International Undertaking on Plant Genetic Resources was adopted here in. Another major step was taken in 1996 with the adoption of the Global Plan of Action at the Leipzig International Technical Conference on Plant Genetic Resources. 

15.1 Sample Objective Question

1. Which one is related to deposition of Micro-organism?
   1. Madrid Agreement
   2. Berne Convention
   3. Budapest Treaty
   4. Patent Cooperation Treaty.
2. WTO was officially commenced in the year...........
   1. 1970
   2. 2003
   3. 2005
   4. 1995
3. Which one is related to protection of Trademarks
   1. Madrid Agreement
   2. Berne Convention
   3. Budapest treaty
   4. None of these
4. One of the four patent office is...........
   1. Patna
   2. Banglore
   3. Varanasi
   4. Mumbai

The Patent Cooperation Treaty provides an overview of an International Treaty which is administered by the World Intellectual Property Organization. The Patent Cooperation Treaty is an international treaty with more than 157 Contracting States by 2022. India deposited its instrument of accession on 7 September 1998 to both international treaties of WIPO in Geneva i.e. Paris convention for the protection of Industrial Property and Patent Cooperation treaty. PCT provides a system for filing a patent application and allow us to obtain patents in multiple countries around the world on the basis of a single patent application.

Quality assurance is fundamental to producing safe, consistent, legally compliant, and commercially successful ice cream. This chapter presents a comprehensive QA framework covering supplier approval, incoming-material inspection, process monitoring, laboratory testing, microbiological control, sanitation, finished-product release, traceability, market quality, and corrective and preventive action. It explains the relationship between quality assurance, quality control, food safety, documentation, and management systems such as HACCP and ISO-based frameworks. The chapter also examines QA department structures, monthly reviews, audits, digitalization, quality culture, and performance indicators. By integrating prevention with measurement and accountability, the chapter demonstrates how effective QA can protect consumers, reduce losses, improve consistency, and strengthen profitability.

This is an international convention governing the recognition of deposits in officially approved culture collections for the purpose of patent applications in any country that is a party to it. Because of the difficulties and on occasion of virtual impossibility of reproducing a microorganism from a description of it in a patent specification, it is essential to deposit a strain in a culture collection centre for testing and examination by others.

This was the year, 1974, the Director General of WIPO convened a committee of Experts to discuss the possibilities of international cooperation over the deposit of microorganisms for patent purposes

Basic knowledge of statistics is important to understand animal breeding analysis. In this chapter, the hypothesis testing, principles of experimental design, types of data and probability distribution are described with examples using appropriate software.

It is known that the phenotype of an individual is a combination of genotype and environment. Animal breeding analysis is mainly done to partition the genetic part from the environmental part so as to assess the actual genetic worth of the animal. In this process correction for the known environmental part is integral to any animal breeding analysis. In this chapter, a list of common factors affecting important economic traits are described. The analytical part for these effects will be explained in later chapters.

Correction or standardization of data is very important in any animal breeding analysis. Unless the candidates under selection are placed on a common platform, various factors affecting the trait will lead to bias in assessing the performance of the individual with respect to a particular trait. For eg. a lamb born during a favorable season, a lactating cow in its third parity, first lactation yield of a cow with longer age at first calving.

Correlations and regression are important components of animal breeding analysis. The genetic correlation is an important genetic parameter that is useful in the planning of breeding programs involving multiple traits. In this chapter, the basic concepts of correlation and regression including method of estimation, assumptions and test of significance are presented. An example using SPSS software is also given at the end for better understanding.

Estimation of variance is primary for any breeding program as all the genetic parameters are dependent on these estimates. The properties of an estimate, their assumptions, type of data, availability of data, and computational requirement make it important to understand the method of estimation of variance components so as to make a wise choice that would suit the breeder’s requirement. This chapter explains the methods of estimation of variance with merits and demerits of each and the reader could use these as a guide to choose the right method of estimation.

The genetic parameters are the basis for any animal breeding program and the important parameters such as heritability, repeatability and genetic correlation are ratios of (co)variances and their reliability depends on the accurate estimation of variance components. Estimates of variance components are used extensively in animal breeding.

The key genetic parameters available with the animal breeder to attain improvement are heritability, repeatability and genetic correlation. This chapter will mainly deal with the theoretical considerations of these parameters.

The goal of an animal breeder is to bring about permanent improvement in different economic traits, either individually (univariate) or simultaneously in multiple traits (multitrait). The phenotypic value available from the animal is a combination of genotypic value and environmental deviation. By permanent improvement in a trait, we mean the improvement in genotypic value, which could be carried over to future generations. The environmental deviations are spatially and temporally specific, and thus will not be the same over generations. The genotypic value, itself is made up of additive effect, dominance deviation and interaction deviation. The additive effect is a value of genes (the sum of the average effect of genes) and hence is more important for the breeder as only the genes are passed on to the next generation.

Domestic livestock plays a major role in meeting the food requirement of human beings. Especially with dwindling resources, improvement in production efficiency is primary to meet the ever-increasing demand for animal protein. The role of an animal breeder is to bring about a sustained improvement in the quality and quantity of such products. Over the years, there has been an enormous improvement in the production ability of livestock and statistical analysis has played a predominant role in the application of principles of animal breeding to improve the productivity of animals. The combination of genetics and statistics developed into a separate branch of science called biometry.

Milk yield is one of the most important economic traits and modelling the lactation curve is important with respect to the unique shape with ascending, persistent and descending phases. The non-linear relationship of daily milk yield with days in milk has led to extensive research in modelling the lactation curve. The nature of the lactation curve, different models, identifying the best model of fit, factors affecting the curve parameters and orthogonal polynomials to model the curve are discussed in this chapter.

Linear models have been extensively used in animal breeding. This chapter explains the basics of linear models including description; fixed, random and mixed models; model diagnostics and covariates. The concept of linear models in animal breeding is explained with representation and actual examples of use in different types of livestock.

Introduction to Linear Models

A model can be defined as a physical, mathematical or otherwise logical representation of a system, entity, phenomenon, or process. Linear models are a way of describing response variables in terms of a linear combination of the predictor variable. They provide a framework for a large set of models whose common goal is to explain or predict a quantitative dependent variable by a set of independent variables that can be categorical or quantitative.

Healthcare represents one of the most promising yet sensitive application domains for Large Language Models. LLMs can process extensive medical literature, electronic health records, clinical narratives and other forms of biomedical information to support diagnosis, research and patient care. This chapter explores the growing role of LLMs in healthcare, beginning with their development from general language models toward specialized biomedical and clinical systems. It examines applications in disease diagnosis, clinical decision support, medical documentation, literature analysis, drug discovery and personalized medicine. LLMs can assist with differential diagnosis, symptom extraction, medical question answering, clinical trial matching and interpretation of medical information. In drug discovery, they can analyze scientific publications, patents and clinical evidence to identify potential drug candidates, targets and opportunities for drug repositioning. Personalized medicine applications involve integrating genetic, lifestyle and treatment-response information to support patient-specific therapeutic decisions. At the same time, the chapter emphasizes major concerns involving patient privacy, hallucination, explainability, regulatory compliance and human oversight. The discussion highlights the importance of responsible integration of LLMs with clinical expertise and established healthcare practices