I research machine learning for wireless networks, a field that sounds narrow until you're in it,
then it touches everything: optimization theory, systems design, signal processing, real-time
decision-making under uncertainty. At Purdue, I work on making Radio Access Networks smarter:
better resource allocation, better resilience, better policies as systems scale.
Before this, IIT Kanpur for signal processing and communication networks, and a stint in embedded systems
writing bare-metal firmware for BLE, Zigbee, LoRa, and 5G.
Outside the lab: ham radio operator (VU2LWH), two-time Google Science Fair Asia-Pacific finalist,
trained Carnatic vocalist and mridangam player, and a guitarist of considerably humbler standing.
Online I go by Aravind Gitamuralee - a surname I made from both my parents' names, a small tribute I've chosen to carry.
InstitutionPurdue University, Dept. of ECE, West Lafayette, IN
Modern O-RAN architectures expose a fundamental tension: the non-RT RIC operates on slow timescales
with global visibility, while the near-RT RIC must react to fast channel dynamics with limited context.
This project develops a cross-timescale, closed-loop reinforcement learning framework coordinating
policy decisions across both layers. At its core is a novel problem formulation that jointly models
channel-distribution-aware resource provisioning and cross-slice SLA risk, treating interference
patterns and traffic demands as distributional signals rather than point estimates.
The goal is slicing policies that are proactively robust to distribution shift, not reactively corrective.
Implemented a custom OpenAI Gym environment for RAN slicing and trained a PPO agent
on real-world Nokia Nordic radio data. Benchmarked against a Proportional Fairness baseline,
tracking per-slice throughput and SLA violation rates across training. Written in Python using
Stable-Baselines3.
Built a discrete-event simulation of a multi-node queuing network to study end-to-end
delay, throughput, and packet loss under varying traffic loads. Implemented in Python
with automated plotting to characterize network behavior under different load conditions.
Analytical and simulation-based performance evaluation of communication network topologies.
Implemented models for throughput, delay, and loss using Newton-Raphson solvers, flow
conservation equations, and multi-topology analysis, all in MATLAB.
A collection of communication engineering implementations spanning MIMO systems and wireless channel
modeling. Covers spatial multiplexing, beamforming, BER analysis under fading channels,
and channel capacity estimation, implemented in Python and MATLAB.
My academic path spans signal processing, communications theory, and machine learning.
Before joining Purdue, I completed a postgraduate degree in Signal Processing and Communication Networks
at IIT Kanpur and worked as an embedded systems engineer, developing firmware for wireless
protocols including BLE, Zigbee, LoRa, and 5G.
