New opportunity

Sr. AI Engineer, Physical AI & Robotics

TELUS Digital · Bengaluru, India

About this role

We are seeking a Senior AI Engineer to own client engagements end to end across Physical AI and robotics, from understanding real-world environment constraints to designing and hands-on building of perception and robotics machine learning solutions through delivery. In this solutions-lead role, you will drive the engagement from problem to production, applying computer vision, perception, and 3D geometry to real deployments.

Responsibilities

Own client engagements end to end including technical discovery, solution architecture, hands-on build, and delivery, acting as the technical authority on feasibility.
Translate ambiguous real-world problems into concrete perception and robotics ML architectures and lead demos and proof-of-concept scoping.
Design and build computer vision and perception systems such as detection, segmentation, and pose estimation on the underlying 3D geometry including SE(3), camera models, projection, and coordinate frames.
Own sensor calibration and fusion including intrinsic/extrinsic, hand-eye, and LiDAR-camera, alongside LiDAR processing like registration, ground segmentation, and object detection.
Design egocentric data collection pipelines using wrist or head-mounted cameras, depth sensors, and IMUs for robot learning.
Bridge simulation and deployment through domain randomization, simulation setup, and sim-to-real gap analysis.
Evaluate, select, and integrate Vision-Language-Action (VLA) models, and fine-tune large pretrained models under data and compute constraints.
Build supporting services such as agentic loops, REST or gRPC APIs, and PostgreSQL to expose model inference and robotics services.
Mentor engineers, drive principled diagnostic decisions, and translate current research into practical client implications.

Required qualifications

Bachelor’s degree in Computer Science, Engineering, or a related field.

Preferred qualifications

Experience with agentic systems and prompt engineering including tool-use agents and multi-step planners.
Background in API development with REST or gRPC, Docker, basic CI/CD, and PostgreSQL.
Familiarity with ROS or ROS2 or equivalent robotics middleware.
Proficiency in C# for Unity-based simulation and tooling.
Experience with fine-tuning large pretrained models under data and compute constraints.
Knowledge of edge or onboard optimization including TensorRT, ONNX, and quantization.
Background in Reinforcement Learning, RLHF, motion planning, kinematics, or URDF.
Hands-on experience with robot hardware such as Franka, UR, or Unitree.
Research orientation including tracking arXiv and forming opinions from first principles.
Publications, open-source work, or notable projects in robot learning or Physical AI.

For full details about this role, click “Apply on company site” to view the employer’s original posting.

Skills for this role

Computer Vision3D GeometryLiDARSensor CalibrationSensor FusionRobot LearningVLA ModelsPythonROSROS2C#UnityTensorRTONNXQuantizationReinforcement LearningMotion PlanningKinematicsURDFAgentic SystemsPrompt EngineeringREST APIsgRPCDockerCI/CDPostgreSQL

Checking your profile…

Preparation suggestions, not the employer’s actual interview questions. Always check the original posting for current requirements.

Understand the role

What matters most

End-to-End Client Engagement Ownership
Drive client interactions from technical discovery and proof-of-concept scoping through solution architecture, hands-on build, and final delivery, serving as the primary technical authority on feasibility.
Perception Systems and 3D Geometry Implementation
Design and build core computer vision and perception capabilities, including detection, segmentation, and pose estimation grounded in underlying 3D geometry such as SE(3), camera models, projection, and coordinate frames.
Sensor Calibration, Fusion, and Data Pipelines
Manage intrinsic, extrinsic, hand-eye, and LiDAR-camera calibration and fusion, process LiDAR data, and design egocentric data collection pipelines using wrist or head-mounted cameras, depth sensors, and IMUs.
Simulation and Advanced Model Integration
Bridge simulation and real-world deployment through domain randomization and sim-to-real gap analysis while evaluating, selecting, and integrating Vision-Language-Action (VLA) and large pretrained models under constraints.

Likely challenges to think through

Bridging the Sim-to-Real Gap
Difficulty in transferring models trained in simulation to unpredictable physical environments. Prepare by demonstrating robust experience with domain randomization, sim setup, and systematic sim-to-real error analysis.
Managing Ambiguous Client Requirements
Translating vague real-world client constraints into concrete perception and robotics ML architectures. Prepare by practicing structured technical discovery, rapid proof-of-concept scoping, and acting as a definitive technical authority on feasibility.
Balancing Edge Constraints with Model Complexity
Integrating complex pretrained or VLA models under strict data and compute limitations. Prepare by reviewing edge optimization techniques such as TensorRT, ONNX, and quantization.

Practice your answers

Technical practice

3D Coordinate Transformations and SE(3)

What to coverExplain how you manage transformations in SE(3) when projecting camera frames to world coordinates. Discuss handling rotation matrices, translation vectors, and coordinate frame alignments during computer vision perception tasks.

Sensor Calibration and LiDAR-Camera Fusion

What to coverDescribe your approach to performing intrinsic and extrinsic calibration alongside hand-eye calibration for multi-sensor setups. Address how you synchronize data streams and handle object detection using fused LiDAR and camera inputs.

Bridging the Sim-to-Real Gap

What to coverWhat techniques do you use to minimize the performance discrepancy between virtual simulation environments and real-world robot deployments? Discuss domain randomization, sensor noise modeling, and sim-to-real gap analysis.

Integration and Fine-Tuning of VLA Models

What to coverHow do you evaluate, select, and fine-tune Vision-Language-Action (VLA) models or large pretrained models under strict compute and data constraints? Explain your strategy for balancing model capacity with onboard execution limits.

Egocentric Data Collection Pipelines

What to coverHow would you design an egocentric data collection pipeline utilizing head or wrist-mounted cameras, depth sensors, and IMUs? Address data synchronization, timestamp alignment, and storage considerations for robot learning.

Edge Optimization with TensorRT and ONNX

What to coverExplain the workflow for optimizing a perception model for onboard edge hardware using ONNX export and TensorRT quantization (e.g., FP16 or INT8). Discuss trade-offs between inference speed and accuracy degradation.

Robotics Middleware with ROS/ROS2

What to coverHow do you structure node communication and handle message serialization within ROS or ROS2 for real-time robotic control? Address latency management and fault tolerance in distributed robotics architectures.

Supporting Services and API Design

What to coverDescribe how you design supporting backend services using REST or gRPC APIs, Docker, and PostgreSQL to expose model inference and robotics telemetry. Highlight considerations for low-latency communication and robust state persistence.

Agentic Systems and Multi-Step Planners

What to coverHow do you implement tool-use agents and multi-step planners for physical AI applications? Discuss error recovery, deterministic constraints, and loop safety when agents interact with physical environments.

Behavioural practice

Managing Ambiguous Client Requirements

What to coverDescribe a situation where you had to translate a vague or ambiguous client problem into a concrete technical architecture and proof-of-concept scope. Use the STAR method to explain how you managed stakeholder expectations and defined feasibility.

Acting as a Technical Authority under Pressure

What to coverTell me about a time you had to make a definitive 'can we actually do this?' feasibility call on a complex engineering project. Share how you diagnosed underlying constraints and communicated your findings to clients or leadership.

Mentoring Engineers and Driving Principled Decisions

What to coverGive an example of how you mentored junior engineers on diagnostic problem-solving—emphasizing diagnosing before coding. Explain how you guided them to translate current research into practical implementation.

Balancing Technical Constraints and Delivery Timelines

What to coverDescribe a project where you faced tight compute, data, or hardware constraints while building a perception or robotics solution. Explain how you prioritized trade-offs to deliver a working system on schedule.

Sharpen your knowledge

Topics to prepare

Computer Vision and 3D Geometry
Review fundamental concepts of 3D geometry including SE(3) transformations, camera models, coordinate frame projections, and perception tasks such as object detection, segmentation, and pose estimation. Exercise: Derive the camera projection matrix and transformation coordinates for a stereo camera setup. Relevance: Directly applies to designing and building robust CV and perception systems mentioned in the posting.
Sensor Calibration and Fusion
Study intrinsic and extrinsic calibration techniques, hand-eye calibration, LiDAR-camera sensor fusion, and LiDAR data processing like point cloud registration and ground segmentation. Exercise: Outline a step-by-step calibration workflow combining a depth camera with a 3D LiDAR. Relevance: Essential for owning sensor calibration and fusion responsibilities in real robotics deployments.
Simulation-to-Reality Gap Analysis
Understand domain randomization, simulation setup, and methods to diagnose and mitigate the sim-to-real performance gap in robotics. Exercise: Design a simulation testing protocol that introduces noise and randomized lighting to evaluate policy robustness before real-world deployment. Relevance: Directly addresses the posting requirement to bridge simulation and physical deployment.

Skills to revise

ROS and ROS2 Middleware
Gain hands-on proficiency with ROS/ROS2 for robot messaging, node communication, and system integration. Exercise: Build a custom ROS2 node in Python or C++ that subscribes to a simulated camera topic, processes bounding boxes, and publishes control commands. Relevance: Directly supports familiarity with robotics middleware stated in the preferred qualifications.
Edge Optimization with TensorRT and ONNX
Practice converting and optimizing deep learning models for onboard hardware using ONNX runtimes and TensorRT quantization. Exercise: Take a pretrained object detection model, export it to ONNX, and run FP16/INT8 quantization to measure inference latency improvements. Relevance: Directly aligns with onboard optimization requirements for deploying models under compute constraints.
API Development and Backend Services
Build supporting services using Python, REST or gRPC APIs, and PostgreSQL to expose model inference and robotics data. Exercise: Create a lightweight gRPC service wrapper around an inference function that logs requests and outputs to a PostgreSQL database. Relevance: Matches the requirement to build supporting services that expose model inference and robotics services.

Your preparation checklist

Prepare a Physical AI and Robotics Project Walkthrough
Compile a structured portfolio presentation detailing an end-to-end robotics or computer vision project you built, emphasizing perception systems, 3D geometry, sensor fusion, or sim-to-real deployment challenges.
Design a Perception and Sensor Calibration Architecture Exercise
Draft a whitepaper or architectural diagram addressing intrinsic/extrinsic sensor calibration, LiDAR-camera fusion, and coordinate frame transformations (SE(3)) to demonstrate your ability to solve complex spatial problems.
Build a Sample Edge-Optimized Inference and API Service
Develop a lightweight containerized service using Python, REST/gRPC APIs, and PostgreSQL that integrates an ONNX or TensorRT quantized model, showcasing your ability to build production-ready supporting infrastructure.
Draft Evidence for VLA Models and Robot Learning Experience
Prepare concrete code snippets, research notes, or documentation illustrating your experience fine-tuning large pretrained models or Vision-Language-Action (VLA) models under compute constraints.
Formulate Insightful Employer Questions on Client Engagements
Prepare thoughtful questions regarding TELUS Digital's typical client discovery process, hardware stack constraints (such as Franka or Unitree robots), and how engineering teams bridge simulation and real-world deployments.

Availability has not yet been verified. Status can change after a check; confirm on the employer’s website. Report an issue · Our checking policy

KEEP EXPLORING

Similar AI jobs

Related skills and specializations in India. Matched to this role, not your profile.

Explore more jobs

Finding similar opportunities…