Communications Engineering and Machine Learning Systems

Purbesh Mitra

240-***-**** # ******@***.*** ï linkedin.com/in/purbeshmitra Website Google Scholar Summary

Wireless systems engineer with hands-on experience in AI/ML for radio technologies, signal processing, and distributed network protocol design. Built and shipped deep learning models for WiFi-8 (IEEE 802.11) with edge deployment at MediaTek. Leading agentic AI systems for RF spectrum analytics with US Army Research Lab. Strong foundation in information theory, optimization, and wireless communications with 11 IEEE publications. Experience

US Army Research Lab – University of Maryland 2024 – Present Technical Lead, AI-Native Wireless Systems Adelphi, MD

• Leading end-to-end development of agentic AI system for RF spectrum data applications: built LLM-based agents integrated with domain-specific tools for real-time RF analytics, anomaly detection, and interference identification.

• Designed evaluation pipelines for agent outputs on real-world RF spectrum data including LTE band monitoring and unauthorized signal detection. Built retrieval tools enabling army personnel to query RF data via natural language. Supervising junior engineers.

MediaTek USA May 2024 – August 2024

Research Intern, Software R&D Division San Jose, CA

• Built deep learning coordination models for WiFi-8 (IEEE 802.11) multi-access point environments using RSSI and CSI data; scaled system to handle 2 orders of magnitude in action-spatial configurations.

• Designed model compression pipeline using knowledge distillation and quantization; achieved 16 model size reduction with 1% performance loss. Delivered compressed model to product team for edge device deployment. Selected Engineering Projects

Distributed Edge Inference System PyTorch [GitHub] 2024 – 2025

• Built distributed inference framework enabling collaborative LLM generation across multiple wireless edge devices with bounded latency guarantees. Implemented accuracy-latency profiling under resource-constrained environments. Published at IEEE PIMRC 2025.

Scalable Network Protocol Design for Gossip Networks Python, MATLAB 2021 – 2024

• Designed and simulated age-aware multiple access protocols achieving O(1) timeliness scaling regardless of network size. Implemented Bayesian optimization for fair resource allocation in heterogeneous topologies. Published in IEEE INFOCOM 2023 and IEEE Trans. Communications 2024.

Decentralized Learning over Wireless Networks Python, PyTorch 2023 – 2025

• Implemented scale-robust algorithms for asynchronous federated learning over gossip networks, optimizing both communication and computational overheads for distributed training across edge devices. Published in WiOpt 2023 and IEEE SPAWC 2024.

Channel Estimation for Non-Gaussian Wireless Channels MATLAB 2018 – 2020

• Developed data-aided channel estimation techniques for systems with additive Laplace noise (low-frequency radar, UWB). Derived Cram er-Rao lower bounds and evaluated antenna diversity and STBC improvements. LLM Post-Training, Evaluation & Application Systems PyTorch, vLLM [GitHub] 2024 – 2025

• Built end-to-end LLM fine-tuning pipelines (LoRA, GRPO) with automated data curation, reducing data requirements by 85–87%. Developed contamination-resistant evaluation framework benchmarked across 30+ models. Prototyped RAG and agent workflows with tool calling. Shipped trained models and datasets on HuggingFace Hub. Education

University of Maryland, College Park 2020 – Present Ph.D. in Electrical and Computer Engineering College Park, MD

• Focus: AI-native wireless, information theory, distributed systems. 11 peer-reviewed IEEE publications. Indian Institute of Technology Delhi 2018 – 2020

M.Tech. in Communications Engineering New Delhi, India Jadavpur University 2014 – 2018

B.E. in Electronics and Telecommunication Engineering Kolkata, India Technical Skills

Wireless & Communications: Information theory, channel estimation, age of information, gossip protocols, wireless sensing

(RSSI, CSI), STBC, antenna diversity, physical layer security, 5G/6G systems, WiFi (IEEE 802.11), UWB, RF spectrum analytics

AI/ML for Wireless: PyTorch, deep learning for wireless systems, knowledge distillation, quantization, Bayesian optimization, federated learning, neural network model design, edge-deployed AI, agentic AI, LLM-based tool calling Programming & Tools: Python, C, MATLAB, Simulink, Linux, Git, Docker, HPC clusters, embedded programming

(Arduino)

Publications (Full list on Google Scholar)

ASUMAN: Age Sense Updating Multiple Access in Networks

• P. Mitra, S. Ulukus, IEEE Allerton Conference, 2022 [Link] Timely Opportunistic Gossiping in Dense Networks

• P. Mitra, S. Ulukus, IEEE INFOCOM, 2023 [Link]

Timely Asynchronous Hierarchical Federated Learning: Age of Convergence

• P. Mitra, S. Ulukus, WiOpt, 2023 [Link]

Age-Aware Gossiping in Network Topologies

• P. Mitra, S. Ulukus, IEEE Trans. Communications, 2024 [Link] A Learning Based Scheme for Fair Timeliness in Sparse Gossip Networks

• P. Mitra, S. Ulukus, IEEE ICMLCN, 2024 [Link]

Scale-Robust Timely Asynchronous Decentralized Learning

• P. Mitra, S. Ulukus, IEEE SPAWC, 2024 [Link]

Age of Information in Gossip Networks: A Friendly Introduction and Literature Survey

• P. Kaswan, P. Mitra, A. Srivastava, S. Ulukus, IEEE Trans. Communications, 2025 [Link] 6G at 1/6g: A Review and Future Directions for Cislunar Communications

• S. Liyanaarachchi, S. Mitrolaris, P. Mitra, S. Ulukus, IEEE Communications Magazine, 2025 [Link] Distributed Mixture-of-Agents for Edge Inference with Large Language Models

• P. Mitra, P. Kaswan, S. Ulukus, IEEE PIMRC, 2025 [Link] MOTIF: Modular Thinking via Reinforcement Fine-tuning in LLMs

• P. Mitra, S. Ulukus, arXiv:2507.02851, 2025 [Link] Semantic Soft Bootstrapping: Long Context Reasoning in LLMs without RL

• P. Mitra, S. Ulukus, arXiv:2512.05105, 2025 [Link]

Contact this candidate