arXiv subjects

Pallab Datta

Publications and source records attributed to Pallab Datta.

4 recordsLinked to original sources

arXiv2025

A Scalable NorthPole System with End-to-End Vertical Integration for Low-Latency and Energy-Efficient LLM Inference

A vertically integrated, end-to-end, research prototype system combines 288 NorthPole neural inference accelerator cards, offline training algorithms, a high-performance runtime stack, and a containerized inference pipeline to deliver a scalable and efficient cloud inference service. The system delivers 115 peta-ops at 4-bit integer precision and 3.7 PB/s of memory bandwidth across 18 2U servers, while consuming only 30 kW of power and weighing 730 kg in a 0.67 m^2 42U rack footprint. The system can run 3 simultaneous instances of the 8-billion-parameter open-source IBM Granite-3.3-8b-instruct model at 2,048 context length with 28 simultaneous users and a per-user inter-token latency of 2.8 ms. The system is scalable, modular, and reconfigurable, supporting various model sizes and context lengths, and is ideal for deploying agentic workflows for enterprise AI applications in existing data center (cloud, on-prem) environments. For example, the system can support 18 instances of a 3-billion-parameter model or a single instance of a 70-billion-parameter model.

cs.DC↗

arXiv2021

Cell-Laden alginate biomaterial modelling using three-dimensional (3D) microscale finite element technique

A novel modelling technique using finite element analysis to mimic the mechanoresponse of cell-laden biomaterial is proposed for the use in bioinks and other tissue engineering applications. Here a hydrogel-based composite biomaterial specimen was used consisting of 5% (V/V) HeLa cells added to alginate solution (4% W/V) and another specimen with no living cell present in alginate solution (4% W/V). Tensile test experiments were performed on both the specimens with a load cell of 25 N. The specimens were bioprinted using an in-house developed three-dimensional (3D) bioprinter. To allow for the nonlinear hyperelastic behavior of the specimen, the specimens were loaded very slowly, at rates of 0.1 mm/min and 0.5 mm/min, during the tensile test. The microscale finite element models developed in Ansys were loaded with similar load rates and their responses were recorded. Both the model results were validated with the experiment results. A very good agreement between the finite element model and the tensile test experiment was observed under the same mechanical stimuli. Hence, the study reveals that bioprinted scaffold can be virtually modeled to obtain its mechanical characteristics beforehand.

cond-mat.mtrl-sci↗

arXiv2021

A finite element analysis model to predict and optimize the mechanical behaviour of bioprinted scaffolds

Bioprinting is an enabling biofabrication technique to create heterogeneous tissue constructs according to patient-specific geometries and compositions. Optimization of bioinks as per requirements for specific tissue applications is a critical exercise in ensuring clinical translation of the bioprinting technologies. Most notably, optimum hydrogel polymer concentrations are required to ensure adequate mechanical properties of bioprinted constructs without causing significant shear stresses on cells. However, experimental iterations are often tedious for optimizing the bioink properties. In this work, a finite element modelling approach has been undertaken to determine the effect of different bioink parameters like composition, concentration on the range of stresses being experienced by the cells in a bioprinting process. The stress distribution of the cells at different parts of the constructs has also been modelled. It is found that both bioink chemical compositions and stoichiometric concentrations can substantially alter the stress effects experienced by the cells. Similarly, concentrated regions of soft cells near the pore regions were found to increase stress concentrations by almost three times compared to the Von-Mises stress generated around the region of cells away from the pores. The study outlines the importance of finite element models in the rapid development of bioinks.

cond-mat.mtrl-sci↗

arXiv2016

Convolutional Networks for Fast, Energy-Efficient Neuromorphic Computing

Deep networks are now able to achieve human-level performance on a broad spectrum of recognition tasks. Independently, neuromorphic computing has now demonstrated unprecedented energy-efficiency through a new chip architecture based on spiking neurons, low precision synapses, and a scalable communication network. Here, we demonstrate that neuromorphic computing, despite its novel architectural primitives, can implement deep convolution networks that i) approach state-of-the-art classification accuracy across 8 standard datasets, encompassing vision and speech, ii) perform inference while preserving the hardware's underlying energy-efficiency and high throughput, running on the aforementioned datasets at between 1200 and 2600 frames per second and using between 25 and 275 mW (effectively > 6000 frames / sec / W) and iii) can be specified and trained using backpropagation with the same ease-of-use as contemporary deep learning. For the first time, the algorithmic power of deep learning can be merged with the efficiency of neuromorphic processors, bringing the promise of embedded, intelligent, brain-inspired computing one step closer.

cs.NE↗