Shivansh Ahuja

§01

summary

Systems-oriented developer focused on C++, distributed training infrastructure, and GPU-backed ML platforms. Built end-to-end orchestration, checkpointing, runtime launch, observability, and managed Kubernetes flows using Python, Go, Kubernetes, and Volcano. Background also includes Java backend development and graphics/game programming.

§02

experience

• Research Commons

  · C++ Developer Intern

  |—— May 2025 - Present ——|   Remote

  [a] C++ / ML Systems [b] Distributed Training Platform [c] Cloud / Observability read deep dive →

• SAP Labs India

  · Scholar

  |—— Aug 2023 - Jul 2024 ——|   Bengaluru, India

  Developed backend APIs for payroll systems in SAP SuccessFactors using Java, Spring Boot, J2EE, and SAP S/4HANA.

• Indian Institute of Technology Delhi

  · Game Developer Intern

  |—— Dec 2021 - Jul 2022 ——|   Delhi, India

  Built a VR pit simulation in Unity and contributed to additional simulation work during the internship.

§03

personal projects

• §01

  » github

  MegaMario

  C++ SFML

  Built a Mario-style platformer in a personal game engine with ECS, factory-based entity creation, collision handling, scene management, and animation control.

• §02

  » github

  OpenGL Graphics Engine

  C++ OpenGL

  Built a basic 3D graphics engine using OpenGL, focused on rendering pipeline fundamentals and low-level graphics programming.

see full archive →

§04

skills

Languages

C++, C#, Java, Python, CUDA

Systems / ML Infra

Modern C++, multithreading, OpenMP, SIMD (AVX), memory management, distributed training, checkpointing, Docker, Kubernetes, Volcano, Ray jobs

Backend / Platform

Spring Boot, J2EE, REST APIs, SQL, Kafka, Git, CMake

Graphics / Games

OpenGL, SFML, SDL, Unity 3D, Unreal Engine 5.1, Cocos2D-x

Math / CS

Linear algebra, 3D math, trigonometry, operating systems, computer architecture, digital design

§05

blogs

• 2026.03.19 research commons »

  Single Node Training with GKE and Ray

  End-to-end Ray workflow for a distributed training job on Kubernetes — cluster setup, the KubeRay operator, Ray job manifests, the training script, and the monitoring dashboard.

• 2026.01.29 research commons »

  Functional-Gemma: What is it, and how to fine-tune it to do better?

  A walkthrough of Google's 270M-parameter FunctionGemma for edge function calling — special token format, the full calling lifecycle, docstring conventions, and fine-tuning results on BFCL.

• 2026.01.11 research commons »

  Intro to Neural Networks

  Neural network fundamentals — core components, the training loop via PyTorch, tinygrad's minimalist take, and cppgrad, a C++ neural-network framework being built from the ground up.

• 2025.12.18 research commons »

  PyTorch Overview — From Research to Prod

  Opening chapter of a PyTorch series — the library's philosophy, its three core components, real-world use cases where PyTorch shines, and setting up your environment to start coding.

see all posts →

§06

achievements

• ✦ BYOG Game Jam 2021 and 2022: reached top 5 in popularity.
• ✦ Brackeys Game Jam 2021: ranked in the top 500 out of 2200 entries.

§07

education

Bachelor of Computer Applications · New Delhi, India

§08

showreel