Abstract: In this talk, I will present a high-level overview of current research efforts in security and cryptography within IBM Research, both globally and in India. I will then present a brief summary of my recent works at the intersection of cryptography and complexity theory, with applications in research and standardization efforts for quantum-safe cryptography. I will conclude by sharing some insights, challenges and experiences from my journey as a researcher so far, with hopefully some useful takeaways for PhD students and early-stage postdocs. The talk will assume no technical background.

Speaker bio: Sikhar Patranabis is a staff research scientist at IBM Research India. His research interests are in theoretical and applied cryptography, with recent focus on quantum-safe cryptography and privacy-preserving analytics. Prior to joining IBM, he was a staff researcher at VISA Research USA and a postdoctoral researcher at ETH Zurich, Switzerland. He received his B.Tech and Ph.D. from IIT Kharagpur. He is a young associate of the Indian National Academy of Engineering (INAE), an ACM India eminent speaker, and the recipient of the President of India gold medal from IIT Kharagpur. His research has appeared in top conferences and journals in security and cryptography, and has been recognized by an IACR best early career paper award at Asiacrypt, an IBM PhD fellowship, and a Qualcomm Research Innovation Fellowship.

Abstract: Just-in-time (JIT) compilers introduce novel optimization opportunities for programs written in modern languages that offer flexibility in writing programs, and my research group is leading the community with cutting-edge research therein. Though perceived as a “tough” area, working with compilers and programming languages, as well as collaborating with real compiler developers, has been an utterly rewarding experience for my students as well as myself.

In this talk, I recount a few important experiences from my research trajectory, a few important decisions that led to more work happen even in constrained times, as well as a few important learnings that would help future researchers expand their limits. Simultaneously, the technical side would include examples to motivate aspiring scientists join the world of improving programs and the programming ecosystem, as well as be ahead of time while fulfilling their career aspirations.

Speaker bio: Manas Thakur is a faculty member in the Department of Computer Science and Engineering at Indian Institute of Technology Bombay. His research interests include program analysis, compiler optimizations, programming languages, software engineering, and parallelization. Manas holds PhD and Masters degrees in Computer Science and Engineering from IIT Madras. His thesis titled “Precise and Efficient Analysis of Java Programs” won an Institute Research Award and the IBM Best Thesis Award in 2019-20. Prior to joining IITB, Manas was at IIT Mandi for three years, wherein he received the Teaching Honour Roll Award in 2021. Manas’s research group at IIT Bombay is named CompL [pronunciation: Compel].

Abstract: Algebraic Complexity Theory studies the complexity of multivariate polynomials. Two of the most important problems in algebraic complexity are – Q1. Algebraic P!= NP [Valiant’79]: to separate the complexity of the symbolic determinant polynomial from that of the permanent polynomial, and Q2. Polynomial Identity Testing (PIT): Design an efficient algorithm to decide whether the polynomial computed by a given polynomial is identically zero or not. GCT (Geometric Complexity Theory) is a novel approach toward proving strong lower bounds in algebraic complexity, via methods from algebraic geometry, and it morally asks the following: Q3. Can approximations help in efficient computation? In this talk, I will discuss Q1-Q3 for a few restrictive models of computations, and their surprising connections to understanding Q1-Q3, in the most general form. The results are part of my PhD thesis, which can be found here: https://www.cse.iitk.ac.in/users/nitin/theses/dutta-2022.pdf

Bio: Pranjal Dutta (https://sites.google.com/view/pduttashomepage) is currently a Postdoc at the School of Computing, NUS, working with Prof. Divesh Aggarwal. His broad research area is Complexity theory; in particular he works on Algebraic Complexity Theory, Geometric Complexity Theory, and Distribution Testing. He finished his PhD in Computer science (2018-2022), from Chennai Mathematical Institute (CMI), under the guidance of Prof. Nitin Saxena (IIT Kanpur). During his PhD, he was supported by the Google PhD Fellowship (2018 -2022). He obtained his bachelor’s in Mathematics and Computer science (2013-2016) and master’s in Computer science (2016-2018) both from CMI.

Abstract: There is a strong incentive to build intelligent machines that can understand and adapt to changes in the visual world without human supervision. While humans and animals learn to perceive the world on their own, almost all state-of-the-art vision systems heavily rely on external supervision from millions of manually annotated training examples. Gathering such large-scale manual annotations for structured vision tasks, such as monocular depth estimation, scene segmentation, human pose estimation, faces several practical limitations. Usually, the annotations are gathered in two broad ways; 1) via specialized instruments (sensors) or laboratory setups, 2) via manual annotations. Both processes have several drawbacks. While human annotations are expensive, scarce, or error-prone; instrument-based annotations are often noisy or limited to specific laboratory environments. Such limitations not only stand as a major bottleneck in our efforts to gather unambiguous ground-truth but also limit the diversity in the collected labeled dataset. This motivates us to develop innovative ways to utilize synthetic environments to create labeled synthetic datasets with noise-free unambiguous ground-truths. However, the performance of models trained on such synthetic data markedly degrades when tested on real-world samples due to input distribution shift (a.k.a. domain shift). Unsupervised domain adaptation (DA) seeks learning techniques that can minimize the domain discrepancy between a labeled source and an unlabeled target. However, it mostly remains unexplored for challenging structured prediction based vision tasks.

Motivated by the above observations, my research focuses on addressing the following key aspects: (1) Developing algorithms that support improved transferability to domain and task shifts, (2) Leveraging inter-entity or cross-modal relationships to develop self-supervised objectives, and (3) Instilling natural priors to constrain the model output within the realm of natural distributions. The main contribution of my research lies in the development of innovative algorithms and frameworks to enhance the transferability of computer vision models in the absence of extensive manual annotations. The research addresses challenges associated with domain shift in structured prediction tasks, proposing unsupervised domain adaptation strategies, unified multi-task frameworks, and source-free domain adaptation settings. Additionally, we introduce novel approaches for self-supervised domain adaptation for monocular depth estimation and 3D human pose estimation, leveraging structural priors and inter-entity relationships. The work also pioneers the effective use of uncertainty estimation for out-of-distribution detection, enabling inference-time self-adaptation in dynamic real-world environments.

Speaker bio:
Jogendra is currently a Research Scientist at Reality Labs Research, Meta, where he is involved in the development of Eye Tracking Technologies for Augmented Reality (AR) and Virtual Reality (VR) applications. Prior to that he completed his PhD at Indian Institute of Science, Bangalore where he was advised by Prof. R. Venkatesh Babu in the Vision & AI Lab (VAL). His research is primarily focused on developing deep learning frameworks that improve the transferability of computer vision models in the absence of manual annotations, specifically addressing the challenges posed by domain-shift. The potential applications of his research span across various industries, including healthcare and AR/VR. In the past, he has worked as an intern at Facebook Reality Labs and also as a Research Assistant with Prof. Aurobindo Rourtray at the Indian Institute of Technology, Kharagpur. He completed his undergraduate studies at the Silicon Institute of Technology, Bhubaneswar.

Abstract:
In times when many socially sensitive decisions are being automated, it is essential to provide solutions with rigorous fairness guarantees. Such an exercise is crucial for its scientific merits and real-world deployments. This talk will highlight recent results that are motivated, in large part, by such considerations and achieve fairness by design under the following two themes:
(i) Collective Welfare: Here, we build on the perspective that, in numerous applications, an algorithm can be viewed as a policy that induces values (rewards) across a population of agents. Hence, we can utilize constructs from mathematical economics (specifically, welfare functions) to quantify the fairness of an algorithm. Such a quantification complements standard metrics (e.g., cumulative regret in online learning), which chase performance efficiency and are not well-justified when fairness is a desideratum. Specifically, I will present recent joint works that extend the notion of regret (in online learning) with a welfarist perspective.

” The limits of efficient computation: an algebraic complexity perspective”

Dr. Nutan Limaye is an Associate Professor in the Computer Science Department at IT University of Copenhagen, Denmark. Before this, she was an Associate Professor in the Department of Computer Science and Engineering, IIT Bombay, India. She finished her Ph.D. at the Institute of Mathematical Sciences (IMSc) Chennai, India in 2009. She did a one-year postdoc stint at TIFR Mumbai, before joining IIT Bombay in 2010.

Her research interests are in theoretical Computer Science. Specifically, she is interested in Algorithms and Complexity Theory. Her recent research has focussed on Algebraic Complexity Theory and Lower Bounds.

She is an Associate Editor for ACM Transactions on Computation Theory (TOCT) and SIAM Journal on Computing (SICOMP). She received the IRCC IIT Bombay Research Publication Award 2019 and the Best Paper award at FOCS 2021. She received the Departmental Award for Excellence in Teaching 2022 for her contributions to teaching activities at IIT Bombay. She has been on program committees of reputed computer science conferences such as STOC, FOCS, ITCS, and ICALP.

She has been a part of the Association of Computing Machinery in many different capacities. Presently, she is the ACM-W Regional Activities Co-Chair. She was associated with ACM India as the vice-chair of the ACM-W from 2019 to 2021. She was also the treasurer for Indian Association for Research in Computing Science (IARCS), India from 2018 to 2020.

Abstract:
Complexity theory attempts to understand the limits of efficient computation. Efficiency is measured in terms of resources such as time or space. And the notion of computation is formalised by specifying the rules of computation. The central question that drives the area is: how hard is it to compute certain important and computationally interesting functions? In this talk, we will approach this question from the algebraic complexity perspective.

Specifically, we will study how hard it is to compute certain polynomials. Polynomials are mathematical objects that arise naturally in many areas of computer science such as coding theory, cryptography, and algorithm design. Valiant in 1979 formalised models of computation of polynomials and started their systematic study. He posed the following fundamental questions.

Over the last decade, my research has provided partial answers to some of these fundamental questions. In this talk, I will present an overview of the work related to these themes over the last couple of decades. I will end with some compelling research directions that emerge from this work.

Chitra Babu is currently serving as the Professor of the Department of Computer Science and Engineering at one of the leading Engineering colleges in Tamilnadu, namely, Sri Sivasubramniya Nadar(SSN) College of Engineering. She earlier served as the Head of the department, for 15 years. She received her Ph.D. from IIT Madras and MS from The Ohio State University, USA. She has worked in AT&T Bell Labs at Columbus, Ohio, USA.
She has published over 75 research papers in refereed international journals and in the proceedings of international conferences. She has graduated 5 Ph.D scholars. Her research interests are in in building data-intensive distributed systems making use of the Software Engineering principles.
She has been serving as a member of the ACM India Education Council for the past 8 years. She has served as Vice Chair of ACM Chennai Professional Chapter during 2017-19. She served as a member of the ACM-W India Council from 2019-2023. She also served as the Chair of iSIGCSE (Special Interest Group of Computer Science Education) Chapter for 2 years from 2021-23.

Ruth Lennon, she/her, is a lecturer with over 20 years of experience in the Department of Computing in Atlantic Technological University Donegal, Ireland. Ruth has been a member of the ACM for over 20 years and is the Global Chair of the ACM-W and a member of the DEI committee. She is the vice-chair of the IEEE UK & Ireland Women in Engineering and a member of the IEEE UK&I DEI committee to promote inclusive strategies for computing. She has worked on projects to support and encourage women in computing and engineering career pathways for as long as she has been involved in computing. Highlighting the vital role that women can play in any technical team is core to her voluntary work.

Dr. Geetha Manjunath is the Founder, CEO and CTO of NIRAMAI Health Analytix, and has led the company to develop a breakthrough AI solution for detecting early stage breast cancer in a non-invasive radiation-free manner. Geetha holds a PhD from IISc and management education from Kellogg’s Chicago. She comes with over 25 years of experience in IT innovation. She has proposed and led multiple AI projects at Xerox Research and Hewlett Packard India. Before starting NIRAMAI, Geetha was a Lab Director for Data Analytics Research at Xerox India. Geetha has received many international and national recognition for her innovations and entrepreneurial work, including CSI Gold Medal, BIRAC WinER Award 2018 and is also on the Forbes List of Top 20 Self-Made Women 2020. She was declared the winner of Women Health Innovation Showcase Asia, Accenture Vahini Innovator of the Year Award from Economic Times and Women Entrepreneur of the Year 2020 by BioSpectrum India. Geetha is also an inventor of 20 US patents and more pending grant.

Dr. Novaline Jacob is Scientist ‘G’ & Deputy Director at Advanced Data Processing Research Institute (ADRIN), Department of Space, Secunderabad. She joined ADRIN in the year 1992 after completing M.Sc in Applied Geology and M.Tech in Remote Sensing from College of Engineering Guindy, Anna University, Chennai. She was Gold Medalist during her under graduation and graduation program. Later obtained Doctoral Degree PhD in Spatial Information Technology from JNTU Hyderabad during the year 2008.

She pioneered to work on Advanced modeling techniques with Remote Sensing & Geospatial data for solving real world problems in varied areas. She has been working in the field of Satellite Image processing, GIS, Data Analytics with image / video /audio / text data, Spatial & dynamic modeling, Knowledge Engineering, Operations Research, Expert Systems, Fuzzy logic, Decision Support Systems, etc for more than 30 years. Current research interest includes, Image understanding using AI/ML techniques for object detection, change detection, Image Super Resolution, NLP, quantum technologies etc.

She has established, developed workable methodologies and led teams for solving problems in diverse civilian, strategic domains for MoD, MHA and spearheaded the projects with various Departments and Ministries. Her contributions towards design and development of Decision Support Systems for various National projects including National Resource Information System, Integrated Resource Information System for Desertic Areas, Landslide & Snow Avalanche Hazard Zonation, System for Operations Planning & Predictive Analytics for Strategic Community to name a few, found practical applications of Remote Sensing/GIS and being used in many parts of the country for planning and decision-making.

She has more than 50 research papers published in International, National Journals and Symposiums.

Dr. Alpana Dubey is principal director at Accenture Labs Bangalore where she heads Digital Experiences R&D group. Her research interests are in the areas at the intersection of HCI, AI, and software engineering. She is member of the board of directors of iHUB Divyasampark, established at IIT Roorkee by the government of India and ACM India Women Council member. She comes with a consistent track record of research and development technologies of significant business impacts. She has been actively contributing to academic bodies such as ACM, IEEE, as conference chair, steering board member, and journal editor. She holds 40 issued / filed patents and 41 peer reviewed papers, in reputed venues. She is senior member of IEEE and holds Ph.D. in Computer Science from IITK.

Abstract: In this talk, I will share insights from my PhD research journey and subsequent faculty career, highlighting key challenges faced and factors that played a key role in overcoming obstacles. The talk aims to offer guidance for students and aspiring academics through practical lessons. I will also briefly discuss recent research projects in our group and available research/career opportunities at IIT Hyderabad.

Speaker’s Bio:
Jyothi Vedurada is an assistant professor at the Dept. of Computer Science and Engineering, IIT Hyderabad. Her research interests are compilers, program analysis, and high-performance computing. Prior to joining IIT Hyderabad, she was a post-doctoral researcher at Microsoft Research Lab, Bangalore. She received her PhD (+M.Tech) from IIT Madras, supported by the TCS PhD fellowship. She received Google Research exploreCSR Award 2021, Distinguished Paper Award at OOPSLA 2021, Distinguished Reviewer Award for OOPSLA 2019 Artifact Evaluation PC, and Institute Research Award for her PhD thesis. She served as Program Vice-Chair for HiPC 2022, as Artifact Evaluation Chair for SAS 2020 and Program Committee member at multiple conferences and journals.

Abstract:
How hard is it to compute a given multivariate polynomial? In this talk, we will motivate this question and the related problem of showing that certain polynomials are hard to compute. We will sketch some proof techniques for proving hardness and see how these proof techniques lead to efficient algorithms for learning arithmetic circuits and for unsupervised learning.

Speaker’s Bio:
Neeraj Kayal is currently a Principal Researcher at the Microsoft Research lab in Bengaluru. Dr. Kayal works in the areas of complexity theory, algorithms, and related areas of theoretical computer science. Neeraj Kayal received his B. Tech and Ph.D. from IIT-Kanpur and has held postdoctoral positions at the Institute for Advanced Study, Princeton and at DIMACS (Rutgers University). Together with Manindra Agrawal and Nitin Saxena, he was part of the team that discovered the first deterministic polynomial time algorithm for primality testing. This work received the Godel Prize (2006) and the Fulkerson Prize (2006). His recent work has been focused on algorithms and lower bounds in algebraic complexity theory for which he and his coauthors received some best paper awards at various conferences. For this body of work, he also received the Infosys Prize (2021) and Bhatnagar Award (2022). Most recently, he has been working on unsupervised learning.