ML Engineer - Simulation (Universal Robots, Bangalore - India)

Summary: Trains, evaluates, and ships perception and 3D generation models for the simulation platform used in robot testing. Builds data, training, and evaluation pipelines so models are reproducible at scale.

Our Purpose

TERADYNE, where experience meets innovation and driving excellence in every connection. We are fueled by creativity and diversity of thought and in our workforce. Our employees are supported to innovate and learn something new every day. We cultivate a culture of inclusion for all employees that respects their individual strengths, views, and experiences. We believe that our differences enable us to be a better team – one that makes better decisions, drives innovation and delivers better business results.

Universal Robots, part of Teradyne Robotics, is the world market leader in collaborative robots (cobots), making products that empower our customers and positively transform the way they work. Teradyne Robotics is a division of Teradyne and includes Universal Robots and Mobile Industrial Robots (MiR). Our global R&D organization of 200+ engineers builds the software and systems behind every cobot we ship.

Opportunity Overview

You will join a team of talented and passionate engineers developing, testing, and optimizing the software and firmware running on our various robot platforms.

As an ML Engineer on the Simulation team, you will train, evaluate, and ship the perception and 3D generation models that power our simulation platform for UR cobots and MiR mobile robots. Simulation is central to how we reduce integration time, validate fleet behaviour, and let customers model their own environments and workflows ahead of deployment — and increasingly, the assets, scenes, and perception capabilities inside those simulations are learned rather than hand-built.

This is a hands-on engineering role at the intersection of computer vision, 3D deep learning, and robotics. You will work closely with our Senior ML Research Scientist to turn research ideas into robust, production-ready models — covering object detection, pose estimation, 3D asset generation, image-to-3D reconstruction, and related tasks — and you will build the data, training, and evaluation pipelines that make them reproducible at scale. You will also collaborate with robotics engineers who integrate your models into the simulation platform, with application engineers who understand real customer deployments, and with the wider R&D organization in Denmark and Bangalore.

Your responsibilities and tasks include:

• Train, fine-tune, and evaluate deep learning models for object detection, instance segmentation, and object and human pose estimation across industrial, manufacturing, and warehouse environments.
• Develop and train models for 3D asset generation and image-to-3D reconstruction, producing simulation-ready meshes, textures, and materials from images, scans, or text descriptions.
• Work alongside the ML team to prototype, reproduce, and productionise state-of-the-art methods — including diffusion models, transformers, NeRF, and Gaussian splatting — taking them from paper to working pipeline.
• Build scalable, reproducible training pipelines across cloud and on-premises GPU clusters, including distributed training, experiment tracking, and model versioning.
• Generate and curate synthetic training data from the simulation platform using domain randomisation and procedural scene variation, and combine it with real-world robot and camera data.
• Develop data collection, annotation, preprocessing, and quality-control pipelines for 2D and 3D datasets, including auto-labelling workflows that scale with simulation throughput.
• Post-process generated 3D assets — mesh cleanup, texturing, collision geometry, and physical properties — so they import reliably into simulation environments (USD, URDF, SDF).
• Define evaluation metrics, benchmarks, and test sets for perception and 3D generation quality, and measure how well models trained on synthetic data transfer to real deployments.
• Optimise models for inference speed and memory footprint (quantisation, ONNX, TensorRT) and package them as services or modules that integrate cleanly into the simulation stack.
• Work with robotics engineers to integrate trained models and generated assets into production simulation workflows and ROS2-based pipelines.
• Write clean, well-tested, and documented code, and participate actively in code reviews to uphold engineering standards across the team.
• Contribute to CI/CD and MLOps tooling that keeps models continuously trained, validated, and deployable.
• Stay current with advances in computer vision, 3D generative AI, and embodied intelligence, and share relevant findings with the team.

All About You

We seek individuals who share our passion and determination. Our commitment to customer success drives us to go the extra mile. If you're ready to join us in this mission, take a closer look at the minimum criteria for the position.

We value diversity in culture, competence, and experience. We expect you to bring a number of qualifications to be considered for this role.

Mandatory Qualifications:

• Bachelor's or Master's degree in Computer Science, Artificial Intelligence, Machine Learning, Electrical Engineering, Robotics, or a related technical field.
• 3+ years of professional experience training and deploying machine learning models, with a significant focus on computer vision.
• Strong programming proficiency in Python, with a demonstrated ability to write production-quality, maintainable code; working knowledge of C++ is an advantage.
• Hands-on experience with PyTorch (or JAX), including building custom models, loss functions, data loaders, and training loops.
• Practical experience training and evaluating models for at least one of: object detection, segmentation, pose estimation, or 3D reconstruction / generation.
• Solid understanding of deep learning and computer vision fundamentals — CNNs and transformers, camera models and projective geometry, coordinate transforms, and standard evaluation metrics.
• Experience building data and training pipelines, including dataset versioning, experiment tracking (e.g. Weights & Biases, MLflow), and reproducible workflows.
• Comfortable working in a Linux environment with GPUs, version control, containerisation, and modern development workflows.

Proficiency required in at least five of the following qualifications:

• 3D deep learning and generative models for 3D content — diffusion-based 3D generation, image-to-3D, multi-view reconstruction, or text-to-3D.
• Neural scene representations such as NeRF, Gaussian splatting, or implicit surface models.
• 6D object pose estimation and tracking, or human pose and keypoint estimation.
• Modern detection and segmentation frameworks and vision foundation models (e.g. YOLO, DETR-family, Detectron2, SAM, Grounding DINO, DINOv2).
• Synthetic data generation and domain randomisation using simulators or rendering engines (Isaac Sim / Replicator, Omniverse, Blender, BlenderProc, Unity).
• 3D geometry processing and asset pipelines (USD, URDF, SDF, mesh processing, Open3D, Trimesh, PyTorch3D).
• Distributed and large-scale GPU training (DDP, FSDP, DeepSpeed, mixed precision) across multi-node clusters.
• Model optimisation and deployment — ONNX, TensorRT, quantisation, and inference serving such as Triton.
• Cloud infrastructure and orchestration — AWS, Azure, or GCP, with Docker and Kubernetes.
• MLOps practices including model registries, automated evaluation, and CI/CD for machine learning.
• Robotics fundamentals and ROS / ROS2, including camera and depth sensor integration.
• Sim-to-real transfer and domain adaptation methods for perception models.
• Publications, open-source contributions, or competition results in computer vision or 3D learning.

Personal Attributes

Candidates should demonstrate ambition, organizational skills, and adaptability. Independence, initiative, responsibility, and ownership are vital. A pragmatic approach, commitment to continuous learning, and dedication to work quality are expected.

What We Offer

At Universal Robots we set a high bar and in return we provide an exciting career in a fast-growing, technology-driven company where you work alongside skilled and talented colleagues from around the world.

• Global team, local impact — Be part of a world-class R&D organisation managed from Denmark while making a tangible difference from our Bangalore office every day.
• Growth and development — A manager with continuous focus on team spirit, employee development, and skill building.
• Empowerment — Ownership of your work and the products you help create; we trust our engineers to make informed decisions that benefit the organisation.
• Health insurance and additional benefits.
• On-site collaboration with co-located product and software teams, fostering strong working relationships and rapid feedback.

*We are only considering candidates local to position location and are unable to provide relocation for this position.*

*This position is not eligible for visa sponsorship.*