Robotics Engineer – Perception Telebortix Innovations | Noida (Delhi-NCR) | On-site | Full-time | 2–6 years About Telebortix Telebortix is a Noida (Delhi-NCR)-based defence-technology startup building autonomy-first unmanned systems for the Indian Armed Forces and dual-use markets. Our stack spans GNSS-denied navigation, an indigenous edge autopilot, sensor fusion, multi-agent swarm coordination and full flight and ground autonomy, fielded on tactical UAVs, heavy-lift VTOL platforms and unmanned ground vehicles. The founding team brings more than a decade of applied autonomy research from Carnegie Mellon University and its National Robotics Engineering Center, across agricultural, aerospace and military programmes. Our systems are in active trials with the Indian Army, and we are scaling from prototype to production. The autonomy stack is the product. The role You will own perception from photons to pose: sensor selection and calibration, real-time pipelines on embedded GPUs, and the outputs the planning and control stack consumes in the field, so a vehicle can see, localise and act when GNSS is denied and the link is jammed. You report directly to the Founder & CEO. What you will do Visual and LiDAR localisation
• Build and harden visual-inertial odometry, LiDAR-inertial odometry and map-based localisation for GNSS-denied flight and ground operation, including long-duration drift management and relocalisation after loss.
• Develop terrain-relative and satellite-map-matching localisation for cross-country flight without GNSS.
• Characterise performance across day, night, dust, low-texture and high-altitude conditions and design fallbacks between sensor modalities. Detection, tracking and scene understanding
• Train, optimise and deploy object detection, segmentation and multi-object tracking models (EO and thermal) for targets, obstacles, landing zones and terrain classes.
• Build visual-servoing and target-lock pipelines that feed real-time guidance under jammed-link conditions.
• Curate datasets from field trials, run active-learning loops and manage labelling, versioning and evaluation. Mapping and 3D perception
• Generate and maintain occupancy grids, ESDFs and 3D maps from depth cameras and LiDAR for obstacle avoidance and landing-site selection.




• Fuse multi-sensor outputs (EO, IR, LiDAR, radar) into a consistent world model for the planner. Embedded deployment and field work
• Optimise pipelines for NVIDIA Jetson-class compute using CUDA, TensorRT and ROS 2; meet latency and power budgets on the vehicle.
• Own intrinsic and extrinsic calibration, time synchronisation and sensor-driver integration on in-house and third-party airframes.
• Run flight-test campaigns at our UP trial site and at Army establishments: define test cards, collect data, analyse logs and close the loop into code.
• Write clean, tested C++/Python; take part in code review, design reviews and sprint planning; support customer demonstrations and technical proposals. What we are looking for We weight evidence of perception systems that ran on a real vehicle, outdoors, repeatedly, far above benchmark scores, coursework or simulation-only work.
• B.
Tech/M.
Tech/MS or PhD in Robotics, Computer Science, Electrical Engineering or a related field.
• 2–6 years in robot perception or computer vision, with at least one pipeline you helped take from prototype to repeated real-world operation on a UAV, UGV or similar platform. Fresh PhD/MS with strong project work also considered.
• Solid foundation in multi-view geometry, camera and LiDAR models, probabilistic estimation and optimisation; comfortable deriving and debugging the maths, not only calling a library.
• Hands-on experience with at least one of: VIO/SLAM on real hardware; LiDAR odometry and mapping on real hardware; deploying detection or segmentation models on embedded GPUs at real-time rates.
• Proficient in modern C++ (C++17 or later) and Python; experience with OpenCV, Eigen, PCL and a deep-learning framework (PyTorch preferred).
• Working knowledge of ROS 2, Linux-based embedded compute and sensor-driver integration (GMSL/USB cameras, LiDAR,



IMU time sync).
• Comfortable in the field: calibrating sensors on a vehicle, reading logs, diagnosing timing and exposure issues and iterating quickly during a test campaign.
• Clear written and spoken communication; able to explain a design trade-off to an engineer and a demonstration outcome to an Army officer. Nice to have
• GNSS-denied navigation in the wild: VINS-Fusion, OpenVINS, ORB-SLAM3, FAST-LIO, LIO-SAM, or your own; terrain
- or satellite-map-relative localisation.
• Factor-graph back-ends (GTSAM, Ceres, g2o) and loop closure / place recognition.
• Thermal (LWIR/MWIR) imaging: calibration, EO–IR fusion and low-light / night operation.
• Multi-sensor calibration tooling (Kalibr, LiDAR–camera extrinsics) and hardware time synchronisation (PTP, trigger lines).
• Model optimisation for the edge: TensorRT, INT8 quantisation, pruning; DeepStream or custom CUDA kernels.
• Learned features and matching (SuperPoint, LightGlue), monocular depth, NeRF/Gaussian-splat or photogrammetry for mapping.
• Radar or event-camera perception; sensor degradation handling in dust, rain and fog.
• Prior work on defence, aerospace or other safety-critical programmes.
• Patents, publications or open-source contributions in SLAM, 3D vision or embedded perception.

What we offer
• Ownership of a core layer of the autonomy stack in a small, senior team, with direct access to the founder and to the end user.
• Your pipelines running on a vehicle in front of the customer within weeks, not quarters; regular field trials rather than a benchmark-only loop.
• Competitive salary benchmarked to the role, with ESOPs for strong performers.
• Learning budget, conference travel and support for patents and publications where programme sensitivity allows.
• Health insurance, paid leave and relocation assistance to Noida where needed. Details
•

Location:
Sector 62, Noida (on-site), with periodic travel to field-trial sites in Uttar Pradesh and to Army establishments across India.
• Eligibility: Indian citizens only (defence programme requirement).
• Start: Immediate; notice periods of up to 60 days accommodated.

📌 Robotics Engineer – Perception (India)
🏢 TelebortiX
📍 India