Ayush
Patra.

Software engineer with a background in Software Defined Networking — two years building low-latency systems for Satellite Communication and ORAN infrastructure at Hughes Systique, recently completed an MSc in Advanced Computer Science at the University of Manchester.

I care about how things work at the lowest level.

At Hughes Systique I worked on Layer 4 for Satellite Communication and ORAN stacks, the kind of work where a 1% improvement in call success rate means something real. I wrote critical code modules in C++ and ran profilers at 2,500 calls per second, and fixed faults I couldn't reproduce in a simulator.

My MSc at Manchester pushed me towards neuromorphic computing and machine learning. I'm equally comfortable in a terminal as I am in a research paper.

I am currently working on a personal CLI tool to support Canonical's networking suite (LXD, MAAS, OVN and K8s), allowing them to supervise kernel network config and obtain resources without clashes.

SDN

LXD, MAAS, OVN, Kubernetes — and the kernel network config underneath all of it.

Systems

Low-latency C++ at Layer 4, runtime profiling at 2,500 calls/second, Satellite and ORAN stacks.

Research

MSc dissertation on neuromorphic routing for NoC architectures. ML applied to network traffic analysis.

Selected work

01

Neuromorphic Routing on Network-on-Chip

MSc dissertation. C++ pipeline comparing fast, accurate, and energy-aware spike routing techniques for neuromorphic platforms. Derived intermediate SNN representations using SNNTorch in Python and benchmarked two routing algorithms at scale.

C++PythonSNNTorchNoCNeuromorphic ComputingSystems
GitHub ↗
02

Intrusion Detection — ML on Network Traffic

Trained a supervised learning model to classify network traffic as benign or malicious using the UNSW-NB15 dataset. Achieved 87% test accuracy with a False Alarm Rate of 0.009.

PythonScikit-LearnMLNetwork SecurityUNSW-NB15
03

netvisor — Canonical Networking CLI

Personal CLI tool for supervising kernel network configuration across Canonical's networking suite — LXD, MAAS, OVN and Kubernetes — allowing resource allocation without clashes. Currently in active development.

GoLinux KernelLXDMAASOVNKubernetesNetworkingCLI
GitHub ↗

More on github.com/ayushpatra11

Experience

Software Developer Engineer

Hughes Systique Corporation · Gurgaon, India

Jul 2022 – Aug 2024
  • Designed an end-to-end low-latency call resource allocation algorithm for Layer 4 in C++, achieving a 97% call success rate and increasing calibration efficiency by ~30%.
  • Enhanced proprietary telecom simulators in Python for heavy call-load scenarios, hitting a 99.1% call completion rate.
  • Conducted runtime profiling, fault analysis, and performance tuning under peak loads of 2,500 calls/second.
  • Built a ReactJS + SQLite3 full-stack ETL application displaying live xApp traffic with 100% accuracy; integrated custom xApps with the OAI 5G core via containerisation.
C++Python5G / OAIORANReactJSSQLite3System DesignUnit TestingGDB / ValgrindFault AnalysisCI / CD

Front-End Developer

SaudaTech · Gurgaon, India

Aug 2021 – Nov 2021
  • Designed mobile and web interfaces for the Broker Mitra App using Figma, improving UX and product design.
  • Implemented OTP login and refined the UI, streamlining the flow for 50+ monthly active users.
ReactHTML / CSSJavaScriptFigmaUI / UXComponent Design

Academic background

The University of Manchester

Master of Science · Advanced Computer Science

Sep 2024 – Sep 2025
  • Grade: 72.5%
  • Manchester, United Kingdom
  • Focus: Networking, Neuromorphic Computing, ML, Distributed Systems
  • GreatUniHackathon — HireAI project

Manipal University Jaipur

Bachelor of Technology · Computer Science Engineering

Jul 2018 – Jun 2022
  • Grade: 88.00%
  • TMA-Pai Merit Scholar
  • ACM Student Chapter, Aperture Photography Club

Heritage Xperiential Learning School

High School Diploma · Science (CBSE)

2004 – 2018
  • Grade: 96.2% — 12th, CBSE
  • Click Photography Club, Cinephilia Videography Club

Tools & technologies

Languages
C++PythonGoTypeScriptJavaScriptBash / ExpectSQL
Networking & Systems
TCP / IPBGPOSPFSDN5G / ORANOVNeBPF (learning)Wireshark
Engineering Practices
System DesignSOLID PrinciplesUnit TestingIntegration TestingCI / CDRuntime ProfilingCode ReviewGDB / Valgrind
Frameworks & Tools
Linux / UnixGit / GitLabReactJSPyTorchLangChainSNNTorchFastAPIPostgreSQL

By the numbers

97%Call success rate — Layer 4 C++ allocation algorithm
99.1%Call completion under 2,500 calls/second load
~30%Efficiency gain from calibration safeguards
87%ML intrusion detection accuracy (UNSW-NB15 dataset)
92.6%CNN image classification accuracy on CIFAR
27
public repos
3
total stars
8
followers

Contribution activity

GitHub contribution graph
LeetCode stats