Built with and by Teycir Ben Soltane•
How to Use•FAQ•GitHub•arXiv.org•
Share:
ArXivCSExplorer
☆☆Bookmarks🏆RSSHow to UseFAQ
Home/Authors/Souvik Kundu

Souvik Kundu

8 indexed papers

Recent (6 mo)
8
With code
0
Influential cites
0
Benchmarked
0

Publications per year

8
26

Top categories

ML×5Architecture×4AI×3Distributed×3Multiagent×2

Frequent co-authors

Tushar Krishna4×
Sarbartha Banerjee3×
Saptarshi Mitra2×
Yifan Zhang2×
Rachid Karami2×
Phyo Pyae Moe Aung2×

Research Timeline

2026
How Far Can Disaggregation Go? A Design-Space Exploration of Attention-FFN Disaggregation for Efficient MoE LLM Serving

The paper systematically analyzes the benefits and limits of Attention-FFN Disaggregation (AFD) for Mixture-of-Experts (MoE) LLM serving, demonstrating that AFD is crucial for achieving high throughput under strict latency constraints.

How Far Can Disaggregation Go? A Design-Space Exploration of Attention-FFN Disaggregation for Efficient MoE LLM Serving

The paper systematically analyzes the benefits and limits of Attention-FFN Disaggregation (AFD) for Mixture-of-Experts (MoE) LLM serving, demonstrating that AFD is crucial for achieving high throughput under strict latency constraints.

SPARQLe: Sub-Precision Activation Representation for Quantized LLM Inference

SPARQLe is a hardware-software co-design framework that exploits the inherent sub-precision sparsity of LLM activations to reduce memory traffic and enable efficient computation on lower-bit datapaths, significantly accelerating inference.

SPARQLe: Sub-Precision Activation Representation for Quantized LLM Inference

SPARQLe is a hardware-software co-design framework that exploits the inherent sub-precision sparsity of LLM activations to reduce memory traffic and enable efficient computation on lower-bit datapaths, significantly accelerating inference.

MOSAIC: Efficient Mixture-of-Agent Scheduling via Adaptive Aggregation and Inference Concurrency

MOSAIC is a novel scheduling framework that significantly accelerates Mixture-of-Agents (MoA) workloads by jointly optimizing expert placement and utilizing confidence-aware adaptive aggregation.

MOSAIC: Efficient Mixture-of-Agent Scheduling via Adaptive Aggregation and Inference Concurrency

MOSAIC is a novel scheduling framework that significantly accelerates Mixture-of-Agents (MoA) workloads by jointly optimizing expert placement and utilizing confidence-aware adaptive aggregation.

A Workflow-Aware Serving Layer for Agentic Applications

The paper presents Dyserve, a workflow-aware serving layer for agentic AI applications that compiles per-node model and verifier choices into an integer linear program, allowing for efficient model selection and verification under serving-time conditions.

TRACE-ROUTER: Task-Consistent and Adaptive Online Routing for Agentic AI

The paper introduces TRACE-Router, a task-level routing framework for large language models that aligns routing with the unit of supervision using delayed task feedback.

Highlighted terms show continued research focus across papers

Papers

cs.AIcs.LGcs.MAEmpiricalRecentJul 24, 2026

TRACE-ROUTER: Task-Consistent and Adaptive Online Routing for Agentic AI

Ritik Raj, Souvik Kundu, Sarbartha Banerjee, Dheemanth Joshi +2 more

The paper introduces TRACE-Router, a task-level routing framework for large language models that aligns routing with the unit of supervision using delayed task feedback.

View →
cs.DCcs.MATheoretical
Recent
Jul 3, 2026

A Workflow-Aware Serving Layer for Agentic Applications

Jiayi Qian, Zishen Wan, Hanchen Yang, Chun Tao +2 more

The paper presents Dyserve, a workflow-aware serving layer for agentic AI applications that compiles per-node model and verifier choices into an integer linear program, allowing for efficient model se…

View →
cs.LGcs.ARRecentJun 2, 2026

MOSAIC: Efficient Mixture-of-Agent Scheduling via Adaptive Aggregation and Inference Concurrency

Saptarshi Mitra, Yifan Zhang, Rachid Karami, Phyo Pyae Moe Aung +4 more

MOSAIC is a novel scheduling framework that significantly accelerates Mixture-of-Agents (MoA) workloads by jointly optimizing expert placement and utilizing confidence-aware adaptive aggregation.

View →
cs.LGcs.ARRecentJun 2, 2026

MOSAIC: Efficient Mixture-of-Agent Scheduling via Adaptive Aggregation and Inference Concurrency

Saptarshi Mitra, Yifan Zhang, Rachid Karami, Phyo Pyae Moe Aung +4 more

MOSAIC is a novel scheduling framework that significantly accelerates Mixture-of-Agents (MoA) workloads by jointly optimizing expert placement and utilizing confidence-aware adaptive aggregation.

View →
cs.ARRecentMay 29, 2026

SPARQLe: Sub-Precision Activation Representation for Quantized LLM Inference

Aradhana Mohan Parvathy, Soumendu Kumar Ghosh, Shamik Kundu, Arnab Raha +3 more

SPARQLe is a hardware-software co-design framework that exploits the inherent sub-precision sparsity of LLM activations to reduce memory traffic and enable efficient computation on lower-bit datapaths…

View →
cs.ARRecentMay 29, 2026

SPARQLe: Sub-Precision Activation Representation for Quantized LLM Inference

Aradhana Mohan Parvathy, Soumendu Kumar Ghosh, Shamik Kundu, Arnab Raha +3 more

SPARQLe is a hardware-software co-design framework that exploits the inherent sub-precision sparsity of LLM activations to reduce memory traffic and enable efficient computation on lower-bit datapaths…

View →
cs.LGcs.AIcs.DCRecentMay 27, 2026

How Far Can Disaggregation Go? A Design-Space Exploration of Attention-FFN Disaggregation for Efficient MoE LLM Serving

Hanjiang Wu, Abhimanyu Rajeshkumar Bambhaniya, Sarbartha Banerjee, Tuhin Khare +8 more

The paper systematically analyzes the benefits and limits of Attention-FFN Disaggregation (AFD) for Mixture-of-Experts (MoE) LLM serving, demonstrating that AFD is crucial for achieving high throughpu…

View →
cs.LGcs.AIcs.DCRecentMay 27, 2026

How Far Can Disaggregation Go? A Design-Space Exploration of Attention-FFN Disaggregation for Efficient MoE LLM Serving

Hanjiang Wu, Abhimanyu Rajeshkumar Bambhaniya, Sarbartha Banerjee, Tuhin Khare +8 more

The paper systematically analyzes the benefits and limits of Attention-FFN Disaggregation (AFD) for Mixture-of-Experts (MoE) LLM serving, demonstrating that AFD is crucial for achieving high throughpu…

View →