Suryateja Devfolio

01. About

I'm a first year postgraduate student at the University of Sydney, pursuing a Master of Computer Science with dual specialisations in Data Science & AI and Cybersecurity.

I spent the better part of last year at ISRO, India's national space agency, building neural models for satellite re-entry prediction and software tools that are now part of daily operational workflows.

I care about writing software that is correct, fast, and purposefully designed. Outside of research, I build independent products. karman, a productivity app with 5 000+ downloads on the Play Store, is one of them.

02. Education

Feb 2026 – Present Current

Master of Computer Science

University of Sydney · Sydney, Australia

Specialisations: Data Science & AI · Cybersecurity

Dec 2021 – Jul 2025

Bachelor of Engineering, Computer Science

Visvesvaraya Technological University · India

GPA 9.18 / 10

Coursework

Data Structures & Algorithms · Operating Systems · Computer Architecture · Computer Networks · Object Oriented Programming

03. Experience

Jul – Dec 2025

Contract Software Engineer

Indian Space Research Organization (ISRO) · Bengaluru, India

Designed an RNN based model for determining the re-entry profile of decaying satellites, achieving results consistently within 5 minutes of standard ISRO in-house methods
Validated model performance during two Inter-Agency Space Debris Coordination Committee (IADC) campaigns, achieving results within a 4.76% difference compared to established internal predictions
Configured a high-availability PostgreSQL cluster ensuring 99.9% uptime for the storage and management of mission-critical logs and operational data
Supported ongoing re-entry analysis operations and contributed to deep space operation support activities

Feb – Jun 2025

Research Intern

Indian Space Research Organization (ISRO) · Bengaluru, India

Built a sequence based deep learning model to estimate the ballistic coefficient of re-entering satellites using historical orbital data
Validated against two real Indian satellites, RISAT-2B and CARTOSAT-2A, with an error margin of 7.5% — within the acceptable threshold for operational use
Model was adopted into routine departmental re-entry analysis reports alongside existing in-house methods

04. Projects

ORBITT

Orbital Retrieval & Behaviour Inspection Tool for TLEs

Designed a graphical tool for real-time satellite data retrieval and ground trace plotting using SGP4 propagation algorithms
Engineered a concurrent processing engine that automated a multi-step manual pipeline, reducing the data retrieval and propagation time for multiple satellites from ~20 minutes to <15 seconds
Deployed a specialised variant of the software currently in active use at ISRO for satellite orbital data inspection

Python SGP4 Orbital Mechanics TLE

Gradient Descent in C++

From scratch ML optimiser at the system level

Implemented vanilla, momentum, and Nesterov gradient descent from scratch in C++, benchmarked across three classic optimisation landscapes: quadratic, Rosenbrock, and Himmelblau functions. Nesterov converged in up to 76% fewer iterations and ran ~8× faster than vanilla gradient descent. Visualisation built in Python.

C++ Python Optimisation ML Fundamentals

karman

Minimalist productivity app · Play Store

A clean, distraction free productivity app built with Flutter. Reached 5 000+ downloads and a 4.7 ★ rating on the Play Store. Featured by Android Authority as "Best App of October 2024".

Flutter Dart Android Play Store

05. Research

Nature Portfolio 14 Feb 2024

Mapping of soil suitability for medicinal plants using machine learning methods

Scientific Reports, Nature Portfolio

Applied machine learning and geographic information systems to identify optimal soil growth zones for 150 vulnerable medicinal herb species. Integrated UN FAO soil datasets with geospatial features and trained a decision tree model to classify 28 distinct subregions, achieving 99.01% soil classification accuracy and 98.76% subregion classification accuracy.

Read paper →

IEEE Xplore 29 Sep 2023

An Early Recommendation Tool to Enhance Medicinal Plant Growth based on GIS and Soil Data

IC2E3-2023 · National Institute of Technology (NIT) Uttarakhand

Used a random forest classifier on soil and geospatial data to identify and recommend cultivation zones for medicinal plants across Karnataka, India. Presented at the IC2E3-2023 international conference at National Institute of Technology, Uttarakhand, supporting conservation and precision agriculture efforts.