Senior Vision Engineer

Role: Senior Vision Engineer

Function: Robotics / Computer Vision / Perception Engineering

Location: Bengaluru, Karnataka, India

Type: Full-time

Industry: Robotics, Autonomous Systems, Industrial Machinery, Construction Technology

About Company

The company builds self-driving electric robots for warehouses and construction sites. It specialises in non-LIDAR autonomous navigation technology.

The flagship product is an electric autonomous robot already deployed in the field. The company is a seed-stage startup based in Bengaluru, founded in 2019, with $3M in funding and a lean team of 25.

It offers a Robot-as-a-Service model, making automation accessible without heavy upfront costs. Backed by notable investors, the company is led by founders with deep expertise in robotics, AI, and engineering.

Position Overview

The company is looking for a Senior Robotics Engineer to own the perception stack on its flagship autonomous robot — from Visual SLAM and sensor fusion to semantic scene understanding and edge-optimised inference. The role sits at the intersection of classical robotics and modern ML, pushing the boundaries of what is possible on embedded hardware in GPS-denied, construction-grade environments. This is a high-ownership position with direct impact on how the robots see and understand the world.

Role & Responsibilities

• Build and optimise multi-monocular Visual SLAM pipelines (ORB-SLAM3, Kimera, OpenVINS) for 360° coverage in GPS-denied indoor and outdoor conditions

• Fuse data from 3D LiDARs, IR/thermal cameras, and mono cameras into a coherent, real-time world model

• Push monocular depth estimation beyond standard baselines using LiDAR-supervised training, multi-view stereo, and learned depth completion

• Build semantic segmentation and 3D scene understanding pipelines for construction-specific classes: scaffolding, formwork, rebar, excavations, workers, and machinery

• Profile and reduce end-to-end latency from sensor capture to costmap output across all perception layers

• Write production-grade C++17 for all real-time, latency-sensitive perception components

• Compile and optimise ML models for embedded hardware using TensorRT, ONNX, and OpenCV

• Prototype Vision-Language-Action (VLA) models — evaluate architectures like OpenVLA and RT-2 derivatives for site-level instruction understanding and navigation action translation

Must Have Criteria

• 4+ years in robotics, autonomous systems, or computer vision (or 3+ years post-PhD)

• Deep hands-on experience with Visual SLAM and VIO — including debugging drift, handling loop closure failures, and tuning on real hardware

• Strong depth estimation experience — monocular, multi-view stereo, or LiDAR-guided approaches

• Semantic segmentation and scene understanding pipeline experience — including segmentation, scene graph construction, and contextual reasoning

• High proficiency in C++17 for real-time systems and Python for ML training and prototyping

• Hands-on experience with PyTorch, ONNX, and TensorRT — from model training through to edge inference deployment

• Solid working knowledge of ROS 2 and DDS/Zenoh middleware for robotics system integration

• Strong fundamentals in linear algebra, probabilistic robotics, state estimation, and Bayes filtering

Nice to Have

• VLA model experimentation on AMRs or robotic manipulators (OpenVLA, RT-2 derivatives)

• Experience with Gaussian Splatting or NeRF for 3D scene reconstruction

• CUDA / GPU kernel optimisation for inference acceleration

• Sim-to-real transfer experience using Isaac Sim or Gazebo

• Publications at ICRA, IROS, CVPR, ECCV, or CoRL

What We Offer

• High-ownership role with direct impact on a deployed autonomous robot product

• Opportunity to work on cutting-edge perception problems in real industrial environments

• Lean, multidisciplinary team with a strong engineering culture and fast iteration cycles

• Opportunity to grow with a well-funded seed-stage robotics startup at an early stage

• Bengaluru-based role with exposure to international deployment environments