Abstract

This thesis is situated in the intersection of History and Natural Language Processing, and examines the historical transformation of trade and pastoralism in the Western Himalayas, while also introducing PastPaths, a domain-adapted Natural Language Processing (NLP) pipeline developed to extract and visualize historical trade networks from unstructured textual archives. The Western Himalayan region, including present-day Kashmir, Ladakh, Himachal Pradesh, and parts of Tibet, has historically functioned as a dense zone of exchange, sustained by the movements of pastoralist communities and transregional trade in commodities such as wool, salt, and grain. Existing historiography has primarily framed the twentieth-century decline of pastoralism as a top-down consequence of colonial policies, often sidelining the agency of local communities. Through a close reading of archival records, census data, and ethnographic literature, this thesis offers a more nuanced account by highlighting how pastoralists strategically adapted to shifting political and economic contexts. These transitions included movement into agriculture, wage labor, and state employment. From a computational perspective, the thesis addresses limitations in historical methodology by developing PastPaths, a pipeline that integrates OCR, Named Entity Recognition (NER), relation extraction, and map generation. The pipeline uses transformer-based models fine-tuned on manually annotated data from the region and is capable of extracting structured information about the trade of commodities. Our results show improved performance over baseline models and demonstrate the feasibility of large-scale trade reconstruction in data-scarce settings. The thesis contributes to both historical and computational research. It reframes dominant narratives about pastoral decline by recovering voices of adaptation and choice, while also showcasing how domain-specific NLP tools can meaningfully support historical inquiry. More broadly, it advances the field of computational humanities by proposing an integrative workflow where historical questions guide model design, and computational outputs inform previous scholarship in the human sciences. This approach offers a framework for bridging disciplinary silos and generating new insights from underexplored textual archives.

Abstract

With major advances in the field of Machine Learning, fully autonomous vehicles are becoming a reality. However, most research towards ML-based approaches for autonomous driving and safety measures so far solely focuses on four-wheelers. Two-wheelers are a primary mode of transport in many developing countries and are also very vulnerable in road accidents due to their open nature. Hence, effective safety measures for two-wheelers can potentially save millions of lives. In this thesis, we first look at the problem of driving event recognition, which is the primary step towards incorporating prediction-based models in safety features. We note that there is no public dataset that we can use as a standard for this task, and hence design a hardware system to collect data. The hardware system is modular and can be attached to any vehicle to collect six-axis acceleration and velocity data at a high polling rate. We tested traditional Machine Learning (ML) models with the annotated dataset for accuracy and find them to be insufficient due to the temporal nature of the data. So, we propose employing Deep Learning (DL) models that are better suited to processing time-series data, such as LSTMs. We also integrate the attention mechanism with the LSTM models to improve their accuracy and performance on long sequences. A comprehensive evaluation of the proposed models reveal that the Bi-LSTM model with attention provides the best accuracy. However, we also investigated the viability of these models by quantizing and deploying them on a Raspberry Pi. Here, the regular LSTM model provides the best efficiency and inference time. However, all the models are viable and successfully run on a resource-constrained platform, thus indicating that there is scope for deployment of predictive models with inference on edge. Next, we introduce a scalable and modular platform for comprehensive two-wheeler riding data collection. The platform is equipped with multiple sensors and high-performance embedded systems, and captures essential data points such as GPS, acceleration, gyroscope, speed, and 360-degree image and depth data which are crucial for developing deep learning models for autonomous navigation and accident prevention. We detail the hardware architecture, consisting of an Nvidia Jetson TX2 NX platform, Raspberry Pi 4 Model B, and various sensors, all tailored to operate efficiently on a two-wheeler. The software methodology is designed to synchronize and process the data effectively, hence allowing us to generate accurate and thorough datasets. Our results demonstrate the potential of the platform towards improving two-wheeler safety and contributing to the advancement of autonomous vehicle technologies. Finally, we provide an analysis of the shortcomings of the platform architecture that we observe after the first few data collection runs, and we detail the fixes that we implement. We also discuss the scope of this work and future work that can be done in the domain of two-wheeler data collection, as well as the deployment of predictive models on two-wheelers for various purposes. Overall, this thesis aims to advance the integration of machine learning and predictive models for two-wheeler safety by addressing critical gaps in data availability, model design, and edge deployment. We demonstrate the potential of lightweight model inference on edge and also develop a novel architecture of comprehensive riding data collection, which is essential for large-scale model training.

Abstract

This thesis explores the intersection of applied natural language processing (NLP) and cognitive neuroscience to address two critical challenges: enhancing knowledge accessibility for low-resource Indian languages and investigating whether computational language models emulate human brain activity during naturalistic language comprehension. Together, these efforts aim to bridge the gap between practical NLP applications and our understanding of how the human brain processes language, with potential implications for designing more inclusive and cognitively aligned AI systems. The first part of this work focuses on addressing the digital divide faced by low-resource Indian languages, which are underrepresented in global knowledge repositories like Wikipedia. To tackle this challenge, we developed a scalable, template-based approach to automatically generate high-quality Wikipedia articles for Telugu, one of India’s widely spoken yet digitally underserved languages. Using structured data sources from platforms such as IMDb, USDA, JSTOR, and IUCN, we extracted and enriched information about diverse domains, including movies, plants, animals, and volcanoes. Advanced translation and transliteration techniques—leveraging tools like Bing Translate API and DeepTranslit—were employed to convert attributes into Telugu, ensuring linguistic accuracy. Manual corrections were applied to handle nuances and ambiguities, resulting in over 25,000 machine-generated articles that adhere to Wikipedia’s quality standards, including word count, formatting, and references. This effort not only expanded the digital footprint of Telugu Wikipedia but also demonstrated the feasibility of automating knowledge creation for low-resource languages. However, the success of this template-based system raised intriguing questions: How do humans process and organize such structured information? Can large language models (LLMs) mimic the cognitive mechanisms underlying naturalistic language comprehension? Motivated by these questions, the second part of the thesis delves into the neural underpinnings of language processing using magnetoencephalography (MEG), a non-invasive neuroimaging technique with millisecond temporal resolution. We conducted experiments on 27 participants listening to naturalistic stories, aiming to investigate how text and speech representations from unimodal and multimodal LLMs align with brain activity. Specifically, we compared embeddings from three multimodal models (CLAP, SpeechT5, Pengi) with those from four unimodal text models (BERT, LLaMA-2, XLNet, FLAN-T5) and three unimodal speech models (Wav2Vec2.0, WavLM, Whisper). Ridge regression was used to predict MEG responses from these embeddings, enabling us to evaluate their ability to capture brain-relevant semantics. To isolate higher-level semantic processing from low-level acoustic features, we employed a residual approach, removing phonological and articulatory cues from multimodal embeddings. Our findings reveal several key insights: First, text embeddings—particularly from multimodal models—closely mirror brain responses during semantic processing, especially beyond 350ms, indicating alignment with higher-order cognitive functions. Second, speech embeddings predominantly capture low-level acoustic features and fail to encode meaningful semantics after 350ms. Third, we uncover an asymmetry in cross-modal knowledge transfer: while text modality benefits significantly from speech-derived information, particularly around 200ms (auditory processing), the reverse is limited. These results deepen our understanding of how multimodal models integrate information across modalities and highlight opportunities to refine their design for applications in low-resource contexts. Together, these contributions bridge applied NLP and cognitive science, offering a novel perspective on designing linguistically inclusive and cognitively aligned AI systems. By juxtaposing the scalability of template-based systems for low-resource languages with the temporal precision of MEG-based brain encoding, this thesis advances inclusivity in digital knowledge and paves the way for future interdisciplinary innovation at the intersection of AI and neuroscience. The findings not only enrich our understanding of how the human brain organizes and processes language but also provide actionable insights for improving the alignment of NLP systems with human cognition. Ultimately, this research underscores the importance of integrating biologically inspired mechanisms into computational models, fostering advancements in both neurolinguistics and applied NLP.

Abstract

People are always in search of viable artificial (man-made) alternatives for all the naturally occurring ones. Such started a quest for artificial muscles, which have a wide range of applications from prosthetics to robotics. For this to happen, we are in dire need of electronics that are bendable, stretchable, etc. This is where the research field of Flexible Electronics pitches in; the focus of the work here is to study a specific type of actuators called the Dielectric Elastomeric Actuators (DEA) and how they can be successfully used in making artificial muscles, even though the making of artificial muscles is far away, many preliminary advancements and tests have to be performed on the technology of interest to make sure it is suitable and can act as a viable replacement. The present focus is to try and construct the basic building blocks required to reach the end goal of achieving fully functional artificial muscles. For the DEA, there are a lot of considerations to be made in the process, starting from the type of materials to be used to the process of building them all. The aspects mentioned above will have to be studied and experimented with to progress towards the end goal. Literature search on DEA’s and the field of flexible electronic products itself was also done to get the gist of the field and for a comprehensive knowledge gain so that all the basic elements and path to be followed remains crystal clear right from the beginning. Along with the substrate and other components, the conductive materials used in these also have to endure the applied strain; hence, thin film-based conductors are being studied extensively. We report the bending behavior of conductive CNT-PDMS thin-film channels. Transmission line method (TLM)-based resistance calculations have been done to precisely analyse the CNT thin film sheet resistance and the contact resistance. This exercise allows us to look at the effectiveness of CNT channels as a conductive path. Also, it helps us understand their behaviour over concentration, path length and strain caused by bending. We report sheet resistance values of 254 Ω/✷ and 1.41 Ω/✷ for thin films of two different CNT concentrations. The contact resistances for these films were 230 Ω and 315 Ω respectively. We observed that bending-induced strain caused the resistance values to increase from 3.95 kΩ for flat to 13.1 kΩ for a bending radius of 1.2 cm. Our work towards artificial muscles is an attempt at biomimetic technologies, an essential part of soft robotics.Researchers have been developing and improving many technologies, from prosthetics to artificial skin. We make a step towards fully functional muscles by presenting the fabrication and characterisation of an artificial muscle flapping wing using dielectric elastomer actuators (DEAs). The actuator was fabricated using a urethane acrylate polymer elastomer with carbon nanotube (CNT)-based electrodes. The 40 mm × 20 mm rectangular actuator was powered through two electrodes at the top. It was subjected to different voltages in the range of 500 - 2500 V and the displacement of the bottom edge was observed. We report the actuator edge displacement to be 20 mm, i.e., 0.5 body lengths (BL) at static 2.5 kV. The dynamic behaviour of the device was also analysed by actuating it with a square wave of various frequencies (0 to 10 Hz). We report the resonant frequency of the actuator to be 4 Hz, with a sub-harmonic at 2 Hz. Few Flapping structures have already been reported, wherein a soft wing is paired with a solid actuator or vice-versa, but a fully soft actuator and wing setup is a rarity. We take on this approach as it would significantly increase the appeal and power density of the structures. In addition to fabrication and mechanical characterisation, the electrical response of a fully soft DEA-based flapping wing actuator was analysed to get an overall picture of the structure, which will help in further optimisations. It was subjected to static DC voltages in the range of 1 to 2 kV and the current and energy drawn by the actuator were observed. We report the average settling current varying from 7 µA to 43 µA and energy draw varying from 220 mJ to 715 mJ. Further tests on a similarly fabricated actuator revealed the possibility of Time Dependent Dielectric breakdown (TDDB) encountered at 2kV after nearly 120 minutes.

Abstract

Radio Frequency (RF) biosensors have emerged as a powerful technology for biochemical detection, offering label-free, non-invasive, and real-time sensing capabilities. These sensors detect changes in the electromagnetic properties of a medium, typically through shifts in resonance or impedance. This thesis primarily focuses on the design and development of RF biosensors based on Interdigitated Capacitors (IDCs), which are highly sensitive to variations in the dielectric properties of the surrounding environment. IDCs consist of multiple interdigitated electrodes forming a planar capacitor, where the electric field interacts strongly with the dielectric material between the fingers. Changes in the dielectric properties caused by biochemical substances lead to measurable shifts in the sensor’s resonant frequency or quality factor, enabling sensitive detection. To further enhance sensor performance, this work explores the integration of IDC with Complementary Split-Ring Resonators (CSRRs). CSRRs are resonant structures exhibiting sharp resonance characteristics that improve both sensitivity and selectivity. By incorporating the IDC within or near the CSRR, the sensor benefits from increased electric field concentration and stronger coupling, resulting in a compact, cost-effective device with superior detection capabilities. The design process for the IDC-CSRR sensor begins with selecting a suitable substrate material, such as Rogers or FR4, chosen for their dielectric properties and low loss. IDC parameters-including finger number, width, length, and spacing-are optimized to tailor capacitance and sensitivity. The CSRR is designed with precise gaps that establish a resonant frequency sensitive to dielectric changes. The integrated IDC-CSRR structure is then simulated using full-wave electromagnetic solvers like ANSYS HFSS to optimize resonance characteristics and quality factor. Finally, the sensor is fabricated and experimentally validated through biochemical detection tests, demonstrating its effectiveness in detecting subtle dielectric variations. This thesis also presents a comprehensive review of existing IDC and CSRR-based biosensors and includes studies on IDC-based biochemical detection using grade 4 Whatman filter paper and PLA-based walls, as well as CSRR-based antenna designs on FR4 substrates. The findings contribute to the advancement of high-performance, compact RF biosensors with potential applications in healthcare, environmental monitoring, and food safety

Abstract

The rapid advancement of automotive technology, particularly in Advanced Driver Assistance Systems (ADAS) and Autonomous Vehicles (AVs), has significantly enhanced vehicle safety and driver comfort. Central to these advancements is the task of lane detection, which enables vehicles to perceive their position on the road accurately. However, deploying robust lane detection systems faces significant hurdles, especially on the resource-constrained edge computing platforms typically found within vehicles, and critically within Electric Vehicles (EVs) where energy conservation is paramount for maximizing driving range. Existing lane detection solutions often present a stark trade-off. Deep Learning (DL) models offer superior accuracy and robustness across diverse conditions but demand substantial computational power and energy. Conversely, traditional Image Processing (IP) algorithms are lightweight and energyefficient but frequently struggle with performance in challenging scenarios like poor lighting or complex road markings. Relying solely on either approach is often suboptimal, leading to either compromised safety due to low accuracy or reduced operational range due to high energy consumption. This necessitates intelligent strategies that can effectively manage the accuracy-resource trade-off in real-time. This thesis addresses these challenges by proposing and evaluating efficient and adaptive frameworks specifically designed for lane detection on resource-constrained vehicular platforms. The contributions aim to improve both the inherent efficiency of lane detection processing and the dynamic management of available detection methods. The first contribution introduces S 2P, a novel two-stage superpixel algorithm designed to enhance traditional IP-based lane detection methods like Hough Transform (HT) and Probabilistic Hough Transform (PHT). By intelligently reducing the image area requiring processing through static and dynamic superpixel mapping, S2P significantly improves the accuracy and recall of these lightweight methods (e.g., improving PHT accuracy from 0.89 to 0.95) while maintaining or even reducing power consumption, making them more viable for low-power edge devices.The second contribution presents RL-ALMS, an adaptive model selection framework employing contextual multi-armed bandits and Thompson sampling. RL-ALMS dynamically chooses between different available lane detection models (ranging from lightweight IP to computationally intensive DL) based on real-time context, including road conditions, system performance history, and crucially, the vehicle’s remaining battery state. It incorporates a novel battery-aware reward function and sustainability constraints to intelligently balance immediate detection accuracy with long-term energy preservation, ensuring journey completion. Experimental results demonstrate RL-ALMS’s effectiveness, achieving high average accuracy (90.3%) comparable to always using a DL model, while preserving significant battery reserves (81.2% after 50km) and adapting robustly to varying conditions, significantly outperforming static high-accuracy or energy-efficient baseline strategies. Both proposed solutions are evaluated using relevant datasets (TuSimple) and hardware platforms (Raspberry Pi, Jetson Nano), demonstrating their practical potential for enhancing lane detection capabilities in modern vehicles.

Abstract

The rapid evolution of semiconductor technologies, marked by the transition to FinFETs and increasingly complex CMOS nodes, has intensified challenges in circuit design. Process variations, aging effects, and stringent performance requirements demand innovative strategies to balance precision, reliability, and computational efficiency. This thesis explores machine learning (ML)-driven methodologies to address these challenges, focusing on scalable optimization frameworks for next-generation integrated circuits. Central to this work is the development of advanced machine learning (ML) models suited for circuit data modeling. A chained architecture is introduced to capture process-induced variations in leakage power, voltage, and current across diverse technology nodes and operating conditions. By integrating temperature and voltage fluctuations into its design, this framework significantly enhances prediction accuracy, enabling robust modeling of nanoscale devices across tech-nodes. Complementing this, a novel automated system leverages ML to streamline analog circuit optimization, reducing manual iterations and simulation overhead while maintaining compliance with designer specifications. The thesis further proposes optimization strategies accelerated by ML to address transistor sizing and aging mitigation. A multi-objective evolutionary algorithm framework is designed to rapidly optimize digital logic cells, achieving substantial improvements in power-delay metrics without reliance on traditional simulator-heavy workflows. For FinFET-based systems, an aging-aware optimization methodology dynamically adjusts device parameters to counteract reliability degradation from NBTI and HCI effects, ensuring sustained performance across operational lifetimes. These approaches collectively bridge the gap between computational efficiency and design robustness. By unifying ML-driven modeling with intelligent optimization, this research advances the integration of automation in electronic design. The proposed frameworks not only enhance circuit reliability and efficiency but also empower designers to navigate the complexities of advanced CMOS and FinFET technologies. These contributions lay a foundation for adaptive, high-performance systems in applications ranging from IoT devices to industrial electronics, underscoring the transformative potential of machine-learning (ML) in modern VLSI design.

Abstract

Path tracing is ubiquitous for photorealistic rendering of various real-world appearances. It follows the principles of light transport adapted from physics, which describe light propagation as a set of integral equations. These equations are stochastically evaluated by tracing light rays in virtual scenes. Such stochastic evaluations with ray-tracing form the bulk of the path tracing algorithm, which is widely used in the industry. Stochastic evaluations in path tracing converge to the correct answer in time that is inversely pro- portional to the square root of the number of iterations. This coupled with the fact that the underlying integrals are often complex and high dimensional results in large compute complexity. Research efforts have thus largely focused on accelerating path tracing by improving the stochastic sampling processes. However, it is interesting to look at efficient analytic approximations by making reasonable assumptions on the nature of these light transport integrals. Such analytic methods have the potential to achieve zero variance at the outset. Practically, they are often used in conjunction with stochastic methods thereby achieving lower variance than the fully stochastic counterparts. The primary focus of this thesis is to develop new (semi-)analytic methods and improve existing ones to accelerate direct lighting computations in path tracing. We base our research on the theory of Linearly Transformed Cosines (LTC) applied for direct lighting from area lights. The LTC method produces plausible renderings by building on the principles of light transport from the ground up and has proved useful for tasks other than real-time rendering We make the following three contributions that either build on LTCs or improve it. We first explore fully-analytic direct lighting for arbitrarily shaped area lights, built on LTCs at the core. Due to assumptions of the LTC method, it can only handle polygonal area lights. Furthermore, rendering shadows with LTCs require stochastic evaluations - our contribution here relaxes these as- sumptions, enabling fully-analytic direct lighting with shadows from an arbitrary shaped area light. We show that our method achieves plausible and noise-free renderings compared to semi-analytic LTCs and ground truth ray-tracing, given equal compute budget. Our second contribution improves the LTC formulation by relaxing an assumption that restricts their usage to isotropic GGX reflection distribution functions. Isotropic GGX restricts the kind of reflectance that can be modeled, ultimately restricting the visual fidelity of rendering with LTCs. We identify the core issues that prevent the use of anisotropic GGX, and propose solutions for each. Our contribution not only benefits (semi-)analytic rendering with LTCs but also other related methods that use LTCs at their core.In our third contribution, we extend the theory of Linearly Transformed Spherical Distributions (LTSDs), which are a superset of and core to LTCs, to work with phase functions. This enables us to analytically compute in-scattered radiance, which we build on to semi-analytically render single scattering. Like the original LTC method, we ground our derivations and formulations on the Volume Rendering Equation (VRE) which paves the way for plausible photorealistic renderings despite the bi- ased nature of our method. In effect, our work enables semi-analytic volumetric rendering with area lights. Another focus of this thesis is to accelerate path tracing computations with neural approximations. Path tracing poses a challenge for (semi-)analytic methods as the underlying integrals become increas- ingly more complex and high-dimensional. In this setting, it is beneficial to use a neural network to ap- proximate parts of these complex integrals for improved efficiency. Previous works have demonstrated the usage of neural networks in this manner, however their main focus was on scenes with relatively shorter light paths. We instead focus on scenes where longer paths are required for accurate rendering. More specifically, we tackle the problem of accelerating multiple scattering computations in hair using neural networks. The contributions of this thesis range from exploring analytic and semi-analytic methods to neural approximations for physically based rendering. Moreover, our contributions advance the state of the art, in terms of efficiency, fidelity and capability. We believe our work can inspire further research in computer graphics and adjacent fields.

Abstract

Integrating traffic cameras with deep learning facilitates real-time multi-camera vehicle path reconstruction, benefiting smart city initiatives such as traffic management and public safety. The high demands on latency, bandwidth, and privacy underscore the necessity of edge computing platforms. However, the continuous operation of compute-intensive deep learning algorithms on round-the-clock camera streams can exceed the compute and power capacities of edge data centers. Therefore, it is critical to strategically deploy and activate camera streams to balance compute demands with performance requirements. Currently, there is a lack of tools to support this planning process. This thesis presents Seer, a comprehensive suite of tools and algorithms that aids traffic planners in optimizing traffic camera placement and deep learning model deployment, considering compute and budgetary constraints. Seer encompasses (1) a graph-based algorithm for efficient camera placement incorporating road network topology, (2) a collaborative algorithm for vehicle path reconstruction that enables the use of less accurate but resource-efficient deep learning models, and (3) a scalable traffic simulation framework derived from open-source projects that model city road networks. Evaluations of Seer on two real-world road networks, each with thousands of intersections and simulated vehicle paths, show a 6.3x reduction in compute requirements. This is achieved through a combination of sparse camera deployment and the use of lightweight models, with an average 55% increase in the Hausdorff distance for the path reconstruction algorithm, translating to an absolute increase of 135 to 195 meters.

Abstract

Understanding 3D structure from sparse visual observations is a foundational capability required for robotic perception. In many real-world settings, objects are observed in arbitrary poses, under occlusion, and from limited viewpoints. Inferring complete 3D shape and appearance from such inputs enables downstream tasks like robotic manipulation, scene understanding, and 3D generation. This thesis explores representation learning methods that operate over both explicit and implicit 3D representations to recover structured geometry from minimal input, enabling category-level generalization across unseen instances in challenging conditions. To enable robust 3D inference in such scenarios, we first focus on the challenge of completing object geometry from partial point cloud observations when objects appear in arbitrary orientations. Traditional shape completion methods assume inputs are aligned to a canonical frame, which limits applicability in robotics where objects may be present in non-canonical poses. To address this, we propose SCARP—a method for Shape Completion in ARbitrary Poses. SCARP disentangles shape and pose using a multi-task formulation; it learns rotation-equivariant features for pose estimation and rotation-invariant features for shape reasoning. The network jointly predicts a canonical full point cloud and its 6D pose. Unlike multi-stage pipelines that depend on external canonicalization, SCARP is a single network that learns to complete shapes directly in the observed pose. We demonstrate SCARP’s utility in robotic grasping by showing that completing shapes before grasp prediction reduces invalid and colliding grasps by over 70%, and outperforms prior methods by 45% on completion metrics across several categories. While point clouds offer a compact and geometric view of 3D structure, they are inherently sparse and do not capture appearance or high-frequency detail. To model richer, dense 3D structure directly from images, we turn to implicit neural fields—specifically, Neural Radiance Fields (NeRFs). We introduce HyP-NeRF, a framework for learning category-level priors over Neural Radiance Fields (NeRFs). Our key insight in HyP-NeRF is to use a hypernetwork to not just parameterize a neural network (NeRF) but also generate learnable input encodings, resulting in learning higher frequency details while reducing the computation cost. The hypernetwork predicts both the NeRF MLP weights as well as parameters of a multi-resolution hash encoding(MRHE) of a NeRF conditioned on an instance code, enabling efficient synthesis of NeRFs across unseen objects in a class. To overcome rendering artifacts and improve fidelity, we propose a denoising and finetuning pipeline that leverages a learned 2D denoiser followed by view-consistent NeRF optimization. This formulation supports diverse downstream tasks, including single-view NeRF generation, text-to-NeRF synthesis via CLIP, and retrieval from real-world images. Experiments on high-resolution (512x512) object datasets show that HyP-NeRF achieves state-of-the-art performance in generalization, compression, and instance retrieval, while maintaining photorealistic quality and fast inference. In summary, this thesis presents methods for learning robust and generalizable 3D structure from sparse inputs. SCARP addresses the challenge of shape completion at arbitrary poses through rotationaware modeling of partial point clouds, while HyP-NeRF develops scalable priors over neural fields for instance-conditioned NeRF generation. Together, these approaches offer complementary pathways toward unified, learnable 3D perception in robotic environments. Moreover, the ideas and methods of representation learning presented in these papers have the scope of being scalable and being integrated in future 3D foundation models which require the sufficient inductive biases to be able to account and compensate for the limited availability of training data.