When Machines Think & Act: Reimagining Engineering Education For The Age Of Agentic & Physical AI | Prof. T. G. Sitharam

As machines learn to think and act, engineering education must prepare graduates not to compete with them but to command them responsibly. Artificial intelligence now renamed as 'super intelligence' is crossing two thresholds at once. Agentic AI, meaning systems that can plan, use software tools and complete multi-step tasks with limited supervision, is entering design, simulation, coding and project workflows. Physical AI is bringing perception, reasoning and action into robots, vehicles, factories, laboratories and infrastructure. Together they will change what engineers do, how they learn and how their competence is judged. Drawing on the National Education Policy (NEP) 2020 and the AICTE Model Curriculum, this article argues that engineering education must move from syllabus completion to demonstrated capability, producing graduates who can specify, supervise, verify and take responsibility for intelligent systems that act in both the digital and the physical world. It outlines the changes needed in curriculum, pedagogy, assessment, laboratories and faculty development, and sets out practical steps by which students, teachers, institutions and regulators can prepare for this transition and contribute to Viksit Bharat 2047.

For most of the last century, the engineer's tools were passive. A slide rule, a drafting board, a CAD package or a simulation program did exactly what it was told, and the intelligence in the loop belonged entirely to the person using it. That assumption is now giving way. The first wave of generative AI produced assistants that could answer questions and draft code. The wave now arriving is different in kind. Agentic systems can break a goal into steps, call software tools, run analyses, check their own results and iterate. Physical AI systems can see, reason and act in the real world through robots, drones, autonomous machines and smart infrastructure. As machines learn to think and act, engineering education must prepare graduates not to compete with them but to command them responsibly. The engineer is moving from operating tools to directing, supervising and certifying systems that act on their own.

Engineering education is central to economic competitiveness, social development and national resilience. Engineers design the infrastructure that connects communities, the energy systems that power economies, the digital platforms that organise modern life, and the technologies that improve health, agriculture, mobility and manufacturing. When the nature of engineering work changes this fundamentally, the education that prepares engineers must change with it. The central question is therefore not simply which subjects a student must study, but which capabilities a graduate must demonstrate when an AI agent can draft the calculation and a robot can carry out the task, and how the curriculum can keep those capabilities relevant throughout a working life.

This article examines how agentic and physical AI are likely to reshape engineering education, what that means for curriculum, pedagogy and assessment within the framework of NEP 2020 and the AICTE Model Curriculum, and how each stakeholder can prepare. A curriculum that does not evolve with technology risks preparing students for yesterday's world; a curriculum that chases every new tool risks preparing them for none. The task is to build engineers who can master the next tool, question its limits and use it for the public good.

Engineering Education in a Changing Global Landscape

The global economy is being reshaped by artificial intelligence, automation, advanced computing, robotics, biotechnology, new materials, clean energy and connected digital infrastructure. The boundaries between traditional branches are becoming less rigid. A civil engineer may use AI-assisted design, digital twins, drones and remote sensing; a mechanical engineer may work with collaborative robots and embedded intelligence; an electrical engineer may combine power electronics, semiconductor devices, energy storage and autonomous grid control; and a computer engineer may need to understand cybersecurity, ethics, sustainability and human behaviour. Many consequential innovations now emerge at the intersection of disciplines.

Engineering challenges are increasingly systems challenges, and intelligent agents are becoming part of those systems. Reliable transport requires civil infrastructure, sensors, software, data analytics and public policy and will soon involve autonomous vehicles negotiating with AI-managed traffic networks. Smart manufacturing combines materials, robotics, industrial networks, cybersecurity and human-centred design. Climate resilience demands engineering, environmental science, economics and community participation. Graduates must therefore build deep disciplinary foundations while developing the ability to collaborate across disciplines and with machines, communicate clearly, work with uncertainty and understand the social consequences of technical decisions.

A degree must remain rigorous, but the curriculum can no longer be a fixed inventory of topics delivered in a prescribed sequence. It should be a living framework of foundational knowledge, transferable skills, domain specialisation and applied experience, reviewed periodically with faculty, students, employers, researchers, alumni and professional bodies.

Understanding the Two Waves: Agentic AI and Physical AI

Agentic AI: from assistants to co-workers

Agentic AI refers to systems that do not merely respond to a prompt but pursue a goal. Given a brief, an agent can plan a sequence of actions, invoke simulation packages, databases or code interpreters, evaluate intermediate results and revise its approach. In engineering practice, this means an agent could take a structural brief, generate design alternatives, run finite element analyses, check them against code provisions and draft a design report. Software agents can already write, test and debug substantial blocks of code; research agents can survey literature and summarise evidence; project agents can track schedules, procurement and compliance. The immediate consequence is that technical output can be produced far faster than it can be reviewed. The value of the engineer shifts towards framing the right problem, specifying constraints precisely, judging whether results are correct and accepting accountability for what is delivered.

Physical AI: intelligence that touches the real world

Physical AI refers to intelligence embodied in machines that sense and act: robots that learn from demonstration and simulation, autonomous vehicles and drones, robotic and self-driving laboratories, adaptive manufacturing cells, intelligent energy systems and assistive medical devices. Unlike purely digital AI, physical AI is bound by the laws of mechanics, thermodynamics and electromagnetics, and its errors have physical consequences for people, property and the environment. It depends on classical engineering disciplines such as dynamics, control, sensors, actuators, power electronics, materials and embedded computing. Far from making core engineering obsolete, physical AI makes it more important, because every intelligent machine must ultimately be designed, built, powered, maintained and made safe.

Convergence: agents that orchestrate machines

The two waves are converging. A digital twin of a bridge, a plant or a power network may soon host an agent that monitors sensor data, predicts deterioration, proposes an intervention and dispatches a robot or drone to carry it out. Laboratories are beginning to combine planning agents with robotic equipment that runs experiments around the clock. In such systems, the engineer must understand the whole loop, from data and models to decisions, actuation and consequences, and must know where to place human oversight.

How Agentic and Physical AI Will Transform Engineering Education

From performing tasks to directing systems. Routine work such as standard calculations, boilerplate code, drafting and documentation will increasingly be delegated to AI agents. Graduates will be valued for their ability to formulate problems, decompose them, set constraints and acceptance criteria, coordinate human and machine contributors, and judge the quality of what comes back. Education must deliberately develop this capability, which may be described as supervisory engineering.

Fundamentals become more important, not less. An engineer cannot verify what she does not understand. When an agent produces a design in minutes, the ability to perform an order-of-magnitude check, recognise a physically implausible result, or reason from first principles becomes the main safeguard against error. Mathematics, physics, mechanics, circuits, thermodynamics and computing must be taught for deep understanding and intuition rather than procedural repetition.

Verification, validation and accountability become core skills. Testing, uncertainty quantification, failure-mode analysis, adversarial checking of AI outputs and maintenance of audit trails should be taught explicitly in every branch. Whatever the tools, a human engineer signs the drawing, certifies the system and answers for its failure.

Laboratories become hybrid and sim-to-real. Students will increasingly design and train systems in simulation and digital twins, then deploy them on real hardware such as robot arms, mobile robots, drones, sensor networks and microgrids. Low-cost robotics kits, remote laboratories and shared regional facilities can make this affordable at scale. The gap between simulation and reality, and what it teaches about modelling assumptions, is itself a valuable lesson.

Learning becomes personalised and agent-supported. AI tutors and laboratory agents can offer every student instant feedback, worked practice and help in debugging, something earlier generations rarely had. The risk is that students outsource the thinking that learning requires. Courses must be designed so that agents support effort rather than replace it, with students explaining, defending and extending what they produce.

Assessment must be redesigned. When an agent can complete a typical take-home assignment, such assignments lose value as evidence of competence. Oral examinations, live demonstrations, design reviews, process logs, version histories and reflective reports will provide more trustworthy evidence of what a student can actually do.

Ethics, safety and law move to the centre. Systems that act autonomously raise questions of human oversight, functional safety, privacy, cybersecurity, bias, liability and standards compliance. These cannot be left to a single course in the final year; they must be part of how every system is designed and evaluated.

The role of faculty evolves. Teachers will spend less time transmitting content and more time designing learning experiences, mentoring projects, challenging assumptions and validating student work. Faculty who learn alongside their students and model responsible AI use will be the most effective.

Shifts in engineering education in the agentic and physical AI era

New Areas and New Responsibilities

Artificial intelligence is changing both what engineers need to learn and how they learn it. AI tools and agents can support coding, simulation, design exploration, literature discovery, data analysis and rapid prototyping. At the same time, they make it even more important for students to understand fundamentals, validate outputs, identify bias, protect data, explain decisions and take responsibility for the systems they build. AI literacy, including the ability to work with and supervise agents, should therefore be a shared capability across all branches, not an elective reserved for computer science.

Curricula should introduce emerging areas through common foundations, branch-specific applications and advanced electives. These include machine learning and generative AI; agentic systems and multi-agent orchestration; robotics, autonomy and physical AI; human-robot collaboration; data engineering and cloud computing; cybersecurity and digital trust; semiconductor design and VLSI; edge AI and embedded systems; IoT and digital twins; quantum technologies; advanced manufacturing and Industry 5.0; electric mobility and batteries; renewable energy, green hydrogen and smart grids; climate-resilient infrastructure; health technologies; advanced materials; space systems; and technology for agriculture and water.

No institution can teach every emerging technology in depth, and not every new topic warrants a permanent core course. The better approach is to protect strong foundations in mathematics, science, computing, engineering principles and design, while using modular electives, micro-credentials, short intensive courses, interdisciplinary studios and industry projects to introduce fast-changing fields. Above all, students should learn how to master new tools and frameworks long after graduation. In the age of agentic and physical AI, the capacity to adapt is itself a core engineering competency.

“The aim is not to produce graduates who know every tool available today, but engineers who can master the next tool, question its limitations and use it to solve meaningful problems.”

NEP 2020: A Framework for Transformation

India's National Education Policy (NEP) 2020 provides a framework well suited to this transformation. Its emphasis on holistic and multidisciplinary education, flexible learning pathways, vocational integration, critical thinking, experiential learning and reduced rote content is directly relevant to an era in which machines handle routine tasks and humans must supply judgement. It supports flexible undergraduate structures, multiple entry and exit options where applicable, and the Academic Bank of Credits as part of a more adaptable higher-education ecosystem.

For engineering institutions, the spirit of NEP 2020 must translate into practice. Students should be able to combine engineering with data science, robotics, entrepreneurship, design, public policy, economics, sustainability or health. Internships, apprenticeships, community projects and industry-linked assignments, including work with real robotic and AI-enabled systems, should be woven into the learning journey. Assessment should test the ability to apply concepts, reason through unfamiliar problems, build and validate solutions and work responsibly, not merely reproduce notes. The policy also invites attention to Indian knowledge systems, constitutional values, ethics, inclusion and environmental responsibility, because engineering decisions, particularly those delegated to autonomous systems, affect people, livelihoods, public safety and nature. A future-ready engineer must be both technically competent and socially responsible.

Transforming Pedagogy: Learning by Doing, Discovering and Collaborating

Experiential learning must move from the margins of the curriculum to its centre. Laboratories, design-build-test cycles, field studies, simulations, internships, maker spaces, hackathons, capstone projects and industry challenges connect theory with practice. A water-quality project, for example, can combine sensors, an autonomous sampling drone, data analytics, environmental engineering, public health and community engagement. A smart-mobility project can integrate transport planning, embedded systems, AI agents, cybersecurity and human factors. A warehouse-automation project can bring together mechanical design, perception, path planning, safety standards and the experience of human workers. Such experiences develop technical judgement because students must confront constraints, failures, trade-offs and the needs of real users.

Self-directed learning is equally important. Students should learn to frame questions, locate credible resources, consult technical documentation, learn from open-source communities and reflect on their progress. With AI agents available, they must also learn when to delegate, how to instruct an agent precisely and how to check what it returns. Faculty remain essential, not as the sole source of information, but as mentors who structure inquiry, challenge assumptions, provide feedback and help students distinguish evidence from plausible-sounding error. AI tutors can support personalised practice, provided their use is accompanied by verification, academic integrity and responsible data practices.

Collaborative learning prepares students for real engineering practice, where solutions are rarely produced alone. Increasingly, teams will include both people and AI agents. Multidisciplinary teams should work with defined roles, shared milestones, peer review and joint accountability. Students should learn to explain technical choices to non-specialists, negotiate trade-offs, manage disagreements and document decisions, including decisions about what was delegated to machines. Teamwork should be assessed explicitly rather than assumed to occur simply because students sit together. Assessment must change too: design portfolios, prototypes, laboratory demonstrations, oral examinations, code reviews, reflective journals and peer assessment should complement conventional examinations. Where generative or agentic AI is permitted, students should disclose its use and demonstrate their own reasoning, testing and validation. Assessment should reward originality, sound judgement and reproducible work, not merely polished output.

How Should the Engineering Curriculum Change?

Define graduate attributes first. Every programme should specify what a graduate must demonstrate: strong fundamentals, problem formulation, design and systems thinking, digital and AI fluency, the ability to specify and supervise intelligent agents and machines, data literacy, communication, teamwork, entrepreneurship, ethical judgement, sustainability awareness and lifelong learning. Course outcomes and assessments should map to these attributes.

Protect fundamentals, reduce overload. Mathematics, sciences, computing and disciplinary core courses provide the tools to understand and verify new technologies. Repetition and low-value memorisation should give way to deeper understanding and open-ended problems. A dynamic curriculum does not keep adding subjects; it decides what to retain, integrate, refresh or retire.

Introduce a layered structure. The first year should build common foundations, digital and AI literacy, communication and design thinking, with early hands-on exposure to programmable robots and sensors. The middle semesters should deepen disciplinary knowledge with interdisciplinary minors and studios in areas such as autonomy, intelligent infrastructure or AI-enabled manufacturing. Later semesters should offer flexible specialisations, research pathways, entrepreneurship and substantial capstone work.

Make practice continuous. Each year should include authentic work, whether a community problem, a laboratory investigation, a prototype or an industrial assignment, with defined outcomes, faculty and industry mentoring, and assessed deliverables. Capstone projects should address real problems and, wherever possible, involve external users or sponsors and real AI-enabled or robotic systems.

Build a renewal mechanism. Departments should review emerging skills and graduate outcomes annually, with industry, alumni, startups and research institutions contributing evidence rather than dictating content. Changes should be piloted, evaluated and then scaled on the basis of student learning, employer feedback, research relevance and inclusion.

Invest in faculty and infrastructure. Faculty need sustained opportunities to learn AI-enabled pedagogy, agentic tools, robotics platforms, interdisciplinary supervision and responsible AI. Reform will fail if workload, laboratories, computing resources and incentives remain unchanged.

Embed ethics, safety and sustainability throughout, covering human oversight of autonomous systems, functional safety, energy and material use, accessibility, privacy, cyber risk, reliability, bias and long-term maintenance, rather than confining them to one compulsory course.

Preparing for the Agentic and Physical AI Era

Transformation on this scale cannot be left to regulators or curriculum committees alone. Each stakeholder has a distinct role, and each can begin now.

For students

• Invest in fundamentals. Treat mathematics, physics and core engineering subjects as the means to check and command AI, not as hurdles to clear; strong fundamentals are what distinguish an engineer from an operator.

• Practise working with agents. Use AI agents regularly for coding, analysis and design exploration, but always verify outputs independently and keep a record of what was delegated and what was checked.

• Get your hands on hardware. Build at least one physical system during the degree, such as a robot, drone, sensor network or embedded controller, to learn how models meet friction, noise, latency and failure in the real world.

• Build evidence of capability. Assemble a portfolio of projects, code, prototypes and reflective notes that demonstrates capability rather than relying on grades alone.

• Cultivate durable human skills. Develop communication, teamwork, ethical reasoning and the habit of learning new tools quickly, since these will outlast any specific technology.

For faculty

• Learn the tools first-hand. Gain working familiarity with agentic tools and robotics platforms relevant to the discipline, and model transparent, responsible use in teaching.

• Redesign assignments. Redesign assignments so that AI supports learning rather than replacing it, emphasising problem framing, justification, testing and critique of AI outputs.

• Rethink assessment. Shift assessment towards oral defence, live demonstration, design reviews and process-based evidence.

• Teach across boundaries. Collaborate across departments and with industry to create interdisciplinary studios and projects in areas such as autonomy, intelligent infrastructure and smart manufacturing.

For institutions

• Create shared AI and robotics facilities. Establish shared agentic and physical AI laboratories with simulation platforms, digital twins, robot arms, mobile robots, drones and edge-computing kits, accessible to all branches.

• Ensure access and clear policies. Provide secure, equitable access to computing and AI tools, with clear policies on academic integrity, data protection and disclosure of AI use.

• Open flexible pathways. Offer minors, micro-credentials and short courses in agentic systems, robotics and autonomy that students from any branch can take.

• Reward faculty effort. Recognise faculty time spent on curriculum redesign, laboratory development and professional learning in workload and promotion criteria.

• Build industry partnerships. Partner with industry, startups and research laboratories for live problems, internships, equipment access and co-supervised capstone projects.

For regulators and industry

• Enable rather than prescribe. Encourage institutions to integrate agentic and physical AI through adaptable frameworks, outcome-based accreditation and support for shared regional facilities.

• Co-create with academia. Contribute problems, mentors, data and equipment, and help define the competencies that industry will need, without narrowing education to immediate job roles.

• Develop shared standards. Support standards, safety guidance and professional norms for autonomous systems that can be reflected in the curriculum.

AICTE Model Curriculum: A National Framework

The AICTE Model Curriculum is an important national reference for engineering and technology programmes. It organises foundational sciences, engineering sciences, professional core courses, electives, humanities and practical work, supporting coherence and comparability across institutions while providing a basis for programme design and quality improvement. Before implementation, institutions should check the applicable version, programme-specific requirements and current regulatory directions. The next step is to treat the Model Curriculum as a platform for continuous renewal rather than a static template, aligned with NEP 2020, applicable AICTE norms and the National Credit Framework, where relevant. Emerging fields such as agentic AI, robotics and physical AI, data science, VLSI, cybersecurity, quantum technologies, clean energy and advanced manufacturing can be integrated through updated content, interdisciplinary pathways and practical learning.

AICTE's wider innovation ecosystem, including Institution's Innovation Councils (IICs), the Smart India Hackathon, AICTE Idea Labs, startup initiatives and faculty development, works best when linked to credit-bearing coursework, mentoring, laboratory access and clear learning outcomes. Idea Labs, in particular, could evolve into hubs for physical AI, where students prototype intelligent machines and test them against real problems. A hackathon should not be an isolated event; it should be an entry point into a sustained pathway from problem discovery to prototype development, validation, intellectual property and possible commercialisation.

A national framework must also balance common expectations with local flexibility, because institutions differ in faculty strength, regional industry, infrastructure and student needs. AICTE can provide adaptable frameworks and quality expectations, while institutions build distinctive strengths in semiconductors, robotics and autonomy, mobility, sustainable infrastructure, healthcare technology, agricultural engineering or AI-enabled manufacturing. The objective is neither uniformity nor fragmentation, but a strong national foundation with room for innovation.

Conclusions and Recommendations

Engineering education is a strategic national asset. In a world shaped by agentic and physical AI, climate change, digital competition and rapid industrial transformation, India needs graduates who can do more than pass examinations. They must be able to design, build, test, improve and responsibly deploy solutions, and to direct and verify intelligent systems that increasingly act on their behalf. The curriculum must shift from syllabus completion to demonstrated capability, from passive instruction to active inquiry, from isolated disciplines to purposeful collaboration, and from operating tools to supervising intelligent systems. Institutions, regulators and industry should:

• Make graduate capabilities and measurable learning outcomes the starting point for every curriculum.

• Preserve rigorous fundamentals while reducing repetitive content to make room for design, inquiry, application and verification.

• Embed AI fluency, including the ability to work with and supervise AI agents, across all branches, with advanced specialisations aligned to institutional strengths.

• Provide every student with hands-on experience of physical AI through robotics, autonomy, sensing and embedded systems, using shared and simulation-to-real laboratories.

• Make experiential, self-directed and collaborative learning a continuous, assessed part of every year.

• Redesign assessment to value reasoning, demonstration and responsible AI use over polished output.

• Strengthen internships, apprenticeships, industry projects, research exposure, entrepreneurship and capstone design.

• Use modular courses and micro-credentials to respond to emerging technologies without destabilising core programmes.

• Update AICTE Model Curriculum implementation through evidence-based review aligned with NEP 2020.

• Invest in faculty development, modern laboratories, AI infrastructure and equitable access.

• Measure success through competence, innovation, research quality, employer confidence and societal impact, not only examination scores or placements.

“A nation's engineering strength will be measured not by how many engineers it graduates, but by how effectively those engineers turn knowledge into trustworthy innovation, productive enterprise and public good.”

Machines that think and act will not make engineers unnecessary; they will make good engineers indispensable. The ultimate purpose of a dynamic engineering curriculum is to prepare learners not merely to enter the future, but to shape it. By translating NEP 2020 into authentic learning, using the AICTE Model Curriculum as an adaptable national foundation and preparing deliberately for agentic and physical AI, India can produce engineers who lead globally competitive industries and advance the vision of Viksit Bharat 2047.

References

All India Council for Technical Education. (n.d.). Model curriculum for undergraduate degree courses in engineering and technology. https://www.aicte.gov.in/

Ministry of Education, Government of India. (2020). National Education Policy 2020.

About the Author: Prof. T. G. Sitharam is the former Chairman of the All India Council for Technical Education (AICTE) and former Director of IIT Guwahati. An eminent academician and civil engineering expert, his work focuses on technical education policy, geotechnical engineering, and sustainable infrastructure development.

Disclaimer: The views and opinions expressed in this article are those of the author(s) and do not necessarily reflect the official policy or position of the publisher. While every effort has been made to ensure the accuracy of the information, the publisher is not responsible for any errors or omissions, or for the results obtained from the use of this information.