Kartik Sachdev — Robotics & AI at NVIDIA

KartikSachdev

Solutions Architect at NVIDIA — building production robotics pipelines for humanoid and physical AI systems across EMEA.

Humanoid Robots Foundation Models Sim-to-Real RL Munich, Germany

About

I'm a Solutions Architect in Robotics & AI at NVIDIA, based in Munich, leading customer engagements across the EMEA region. My work is at the intersection of foundation models, humanoid robotics, and simulation — translating frontier research into deployable systems for customers across Europe.

My path here is not a straight line. Trained as a mechanical engineer in Delhi, I worked at an automotive firm in Tokyo, retrained in robotic systems at RWTH Aachen, picked up Japanese along the way, and built a career spanning hardware intuition, computer vision, reinforcement learning, and physical AI. I work across languages, time zones, and technical disciplines.

🇬🇧 English — Fluent 🇩🇪 German — Intermediate 🇯🇵 Japanese — Intermediate

Experience

From mechanical engineering in Tokyo to physical AI at NVIDIA.

Solutions Architect — Robotics & AI

NVIDIA · Munich, Germany

2023 — Present

Leading customer engagements across EMEA at the intersection of foundation models, humanoid robotics, and simulation. Serves as a certified DLI instructor and GSoC mentor, helping engineers and researchers build on NVIDIA's robotics stack.

Isaac SimIsaac LabFoundation ModelsEMEA

Master Thesis — RL Researcher

Siemens AG · Munich, Germany

Apr – Oct 2023

Investigated how the design of observation spaces affects RL performance in robotic manipulation — a nuanced problem at the heart of making RL practical for real robots. Worked with NVIDIA Isaac Sim and the Franka Emika Panda arm.

Reinforcement LearningIsaac SimFranka PandaPyTorch

Computer Vision Engineer

Siemens AG · Munich, Germany

Apr 2022 – Mar 2023

Built a 6D pose estimation system for a bin-picking robot, owning the full pipeline from synthetic dataset generation in Blender to model training on AWS. Automated data preprocessing and training with CI.

PyTorchOpenCVAWSBlenderCI

Robotic Software Developer — Intern

BMW AG · Munich, Germany

Jul 2021 – Feb 2022

Developed state machine logic for collaborative logistics robots using Python and ROS, built simulation environments in Gazebo and MoveIt, and ran a Docker workshop for 20+ engineers across teams.

ROSPythonDockerGazeboMoveIt

🎓

M.Sc Robotic Systems Engineering

RWTH Aachen University · Aachen, Germany · 2020–2023

One of Europe's top technical universities. Coursework in reinforcement learning, deep learning for vision, autonomous systems, and self-driving. Wrote review papers on Visual Attention Networks and Soft Actor-Critic alongside GSoC research.

📋

Product Management Program

Stanford University · 12-week intensive

Research

Published work at the intersection of robot learning, simulation, and dexterous manipulation.

IEEE 2025

A Pathway Study for Future Humanoid Standards

Kartik Sachdev et al. (70+ robotics experts from industry, academia & regulatory bodies)

IEEE Robotics & Automation Society

A strategic roadmap developed by over 70 experts identifying critical gaps in existing robotics standards for humanoid robots. Addresses three interconnected pillars — Classification, Stability, and Human-Robot Interaction — proposing a coordinated path to certify the safety of human-robot interactions. Current standards designed for fixed industrial arms are insufficient for the unique challenges posed by humanoids.

ResearchGate PDF

CoRL 2025

Sim-to-Real Reinforcement Learning for Vision-Based Dexterous Manipulation on Humanoids

Toru Lin, Kartik Sachdev, Linxi Fan, Jitendra Malik, Yuke Zhu

UC Berkeley · NVIDIA · UT Austin

A practical sim-to-real RL recipe for training humanoid robots on challenging dexterous tasks — grasp-and-reach, box lift, and bimanual handover. Features automated real-to-sim tuning, generalized contact-and-goal rewards, divide-and-conquer policy distillation, and a hybrid object representation with modality-specific augmentation. Achieves high success rates on unseen objects.

arXiv PMLR Project Page

NeurIPS Workshop 2023

Learning Dark Matter Representation from Strong Lensing Images through Self-Supervision

Yashwardhan Deshmukh, Kartik Sachdev, Michael Toomey, Sergei Gleyzer

NeurIPS 2023 Workshop: Machine Learning and the Physical Sciences · ML4Sci / GSoC

Self-supervised learning applied to gravitational lensing images to develop unsupervised embeddings across three dark matter models — ultra-light axions, cold dark matter, and halos without substructure. Implements contrastive learning, BYOL, SimSiam, and DINO using ResNet50 and Vision Transformer architectures, demonstrating that self-supervised methods consistently outperform supervised counterparts on classification and regression tasks.

NeurIPS

Speaking

Speaking at conferences and teaching at industry workshops worldwide.

Upcoming

Workshop March 15, 2026

How to Simulate, Train, Validate, and Deploy an End-to-End Robotics Workflow with NVIDIA Isaac

NVIDIA GTC 2026 · San Jose, CA

Full-day instructor-led deep learning workshop guiding participants through an end-to-end robotic workflow — from simulation-based training to real-world robot deployment.

Session Details

Session March 17, 2026

Build Smarter Robots Using NVIDIA Isaac Sim and Isaac Lab

NVIDIA GTC 2026 · San Jose, CA

A 50-minute session exploring next-generation robot autonomy — sensing, planning, and acting — using the NVIDIA Isaac simulation platform.

Session Details

Talk 2026

Robotik und KI — Revolution und Zukunft der Industrie

Hannover Messe · Hannover, Germany

Speaking on the revolution of industrial robotics and AI, the future of autonomous machines, and the role of humanoid robots in modern manufacturing.

Event Info

Past

Talk Jun 2025

Advancing Physical AI with Omniverse on AWS

AWS Summit · Hamburg, Germany

Co-presented with NVIDIA colleagues on the next wave of Physical AI — covering Isaac Sim on Amazon EC2, scaling robot learning with Isaac Lab and AWS Batch, overcoming data scarcity with Cosmos models on EKS, and advancing humanoid robotics with GR00T.

Session Jun 2025

Bridging Sim-to-Real With Robot Learning Tools and Techniques

GTC Paris · NVIDIA CWEP1123

Connect-with-Experts session on simulation-first robot development — synthetic data generation, policy training, and software-in-the-loop testing with Isaac Sim and Isaac Lab to narrow the sim-to-real gap.

Session Details

Session Jun 12, 2025

Develop World Foundation Models With NVIDIA Cosmos

GTC Paris · NVIDIA CWEP1122

Connect-with-Experts session on NVIDIA Cosmos — a breakthrough platform for synthetic data generation and physical AI development, covering world foundation models, tokenizers, and domain-specific post-training for robotics and autonomous vehicles.

Cosmos on HuggingFace

Talk Feb 20–21, 2025

Humanoid Robotics at NVIDIA

International Humanoid Forum 2025 · Switzerland Innovation Park, Biel/Bienne

Presented NVIDIA's work in humanoid robotics at the premier European forum for humanoid robot research and industry — covering Isaac Sim, Isaac Lab, and sim-to-real pipelines for humanoid manipulation.

Instagram

Live Sessions

Live 2025

NVIDIA Isaac GR00T N1.5 × Hugging Face LeRobot 0.4.0

YouTube Livestream · NVIDIA

Live walkthrough of NVIDIA Isaac GR00T N1.5 with the Hugging Face LeRobot 0.4.0 release — featuring new tools, training workflows, and real-robot demos.

▶ Watch on YouTube

Live 2025

Foundation Models on Jetson for Industrial Manipulation — with Vention

YouTube Livestream · NVIDIA × Vention

Technical deep-dive with the Vention team on how Foundation Models and accelerated computing on the Jetson platform are reshaping the industrial manipulator industry.

▶ Watch on YouTube

Judging

Judge Jun 2025

Europe's Best Robotics Startup — automatica 2025

automatica · Munich, Germany

Served on the expert jury selecting Europe's Best Robotics Startup from six finalists. Winner: Scurid, awarded €5,000 and the title at automatica 2025.

Judge & Mentor Jun 2025

Munich AI Robotics Hackathon — LeRobot Worldwide Hackathon

Hugging Face × NVIDIA · Munich, Germany

Judged solutions, mentored participants, and supported the global robotics community at the Munich leg of the LeRobot Worldwide Hackathon. Several winning teams incorporated NVIDIA Isaac GR00T.

NVIDIA Deep Learning Institute

Certified DLI Instructor

Certified to deliver official NVIDIA DLI courses to engineers, researchers, and enterprise teams:

Rapid Application Development with Large Language Models
NVIDIA Isaac for Accelerated Robotics

KartikSachdev

About

Expertise

Experience

Solutions Architect — Robotics & AI

Master Thesis — RL Researcher

Computer Vision Engineer

Robotic Software Developer — Intern

M.Sc Robotic Systems Engineering

Product Management Program

Research

A Pathway Study for Future Humanoid Standards

Sim-to-Real Reinforcement Learning for Vision-Based Dexterous Manipulation on Humanoids

Learning Dark Matter Representation from Strong Lensing Images through Self-Supervision

Speaking

Upcoming

How to Simulate, Train, Validate, and Deploy an End-to-End Robotics Workflow with NVIDIA Isaac

Build Smarter Robots Using NVIDIA Isaac Sim and Isaac Lab

Robotik und KI — Revolution und Zukunft der Industrie

Past

Advancing Physical AI with Omniverse on AWS

Bridging Sim-to-Real With Robot Learning Tools and Techniques

Develop World Foundation Models With NVIDIA Cosmos

Humanoid Robotics at NVIDIA

Live Sessions

NVIDIA Isaac GR00T N1.5 × Hugging Face LeRobot 0.4.0

Foundation Models on Jetson for Industrial Manipulation — with Vention

Judging

Europe's Best Robotics Startup — automatica 2025

Munich AI Robotics Hackathon — LeRobot Worldwide Hackathon

Certified DLI Instructor

Writing

NVIDIA Isaac GR00T in LeRobot

Generalist Robot Policy Evaluation with Isaac Lab-Arena & LeRobot

Benchmarking Vision Transformers for Dark Matter Substructure Classification

Let's Connect