Senior ADAS Perception Research Engineer · Autonomy & ADAS
Surya Kiran
Cherupally
Building production-grade multi-sensor fusion and real-time autonomy stacks. Currently at CNH Industrial — Scottsdale, AZ.
95%
reduction in safety-critical false positives
2–3×
TensorRT inference speedup on Orin
4.0
GPA · MS Robotics · ASU
About
6 years building perception that ships.
Senior ADAS Perception Research Engineer with 6 years delivering production multi-sensor fusion, 3D scene understanding, and real-time inference on NVIDIA Orin/Xavier. Architected falconeye — an internal camera-LiDAR-4D radar state estimation and multi-object tracking stack — achieving a 95% reduction in safety-critical false positives. Active part-time research/prototyping in BEV transformers and DrivingGaussian neural reconstruction while focusing full-time on SurroundView and falconeye v2 features. MS Robotics GPA 4.0/4.0; formal thesis defense scored 4/4.
Currently at CNH Industrial in Scottsdale, building the next generation of autonomous agricultural equipment — from sensor fusion architecture to real-time edge inference on NVIDIA Orin.
95%
reduction in safety-critical false positives
2–3×
TensorRT inference speedup on Orin
4.0
GPA · MS Robotics · ASU
Experience
Where I've built.
Senior ADAS Perception Research Engineer
- Architected falconeye — internal production C++ multi-object tracking & sensor fusion stack: two-stage pipeline GNN non-Hungarian (LiDAR/radar/stereo) + Hungarian (camera) → GNN cross-modal track fusion. 95% reduction in safety-critical false positives.
- Built GPU-accelerated 3D SurroundView (bowl mesh: flat + torus + cylinder) with multi-camera extrinsics calibration. POC in 2 weeks → 60 Hz GPU / 20 Hz CPU with odometry-based articulation.
- Researching/prototyping BEVFormer part-time — unified multi-camera + LiDAR BEV representation; benchmarked on nuScenes 3D detection vs. late-fusion baselines.
- Deployed TensorRT optimization pipeline (FP16/INT8 quantization, layer fusion) on NVIDIA Orin. Achieved 2–3× inference latency reduction within operational accuracy thresholds.
- Researching/prototyping DrivingGaussian neural scene reconstruction part-time — replace bowl projection with scene-adaptive 3D Gaussian splatting from calibrated 8-camera rig.
- Evaluated 3D/4D radar across 6 vendors; built Safety Test Bench (200+ field scenarios and growing, nightly regression) and internal LUMOS visualization tools for field triage.
Research Assistant
- Built AERO autonomous robot: YOLOv4 object detection (75% AP) + multi-object tracking (90% accuracy) + U-Net semantic segmentation (92% accuracy).
- Engineered multi-sensor fusion (2D LiDAR + ultrasonic + cameras) improving depth estimation accuracy by 30%.
- Reduced computational time by 25% with Hybrid A* algorithm, optimizing pathfinding in complex spaces.
- Built autonomous drone perception — YOLO object detection, optical flow tracking, hand-gesture control via MediaPipe/PID.
Computer Vision Engineer
- Invented lossless spherical-to-image projection for precise fruit color categorization; organization filed patent paperwork before departure.
- Re-engineered YOLOv4 for CPU edge deployment achieving ~40% mAP at 100 fps on standard hardware.
- Designed regression-based anomaly detection system for load cells (MSE 0.0241).
- Built Auto-ML stack for image classification that aided in fine-tuning models 7× faster.
Projects
Things I've built.
Open-source projects spanning autonomous robotics, multi-agent AI, computer vision, and reinforcement learning.
R
Multi-agent AI Research System
Multi-agent AI Research System
Apr 2026 – Present · Side project
ResearchSquid
A side project for researching complex topics using institute-style agent workflows. Program Director agent breaks research questions into agenda items; Scientist agents execute tasks from shared queues; Neo4j stores canonical research artifacts with pgvector retrieval.
Multi-agent · Graph memory · Docker sandbox execution
Gesture & Vision-Controlled UAV
Dec 2021 – Apr 2022 · ASU Research Assistant
Autonomous Tello Drone
Full-stack autonomous drone system on DJI Tello: real-time face detection + autonomous tracking via MediaPipe, body pose estimation (33 3D landmarks) + KNN classifier for gesture commands, YOLOv4 object detection, collision avoidance with PID control, and 3D path planning.
★44 GitHub stars · 7 forks
FK/IK Simulation & Body Control
Aug 2021 – Dec 2021 · ASU · MAE 547
Quadruped Robot
Full kinematic simulation of a 12-DOF quadruped: forward kinematics via DH parameters, inverse kinematics for individual leg control, Jacobian-based velocity computation, body-frame transformation matrices. Interactive keyboard control of x/y/z/roll/pitch/yaw with real-time trajectory generation.
MAE 547 · ASU Robotics
GAIL on CarRacing-v2
Aug 2021 – Dec 2021 · ASU · EEE598
Inverse Reinforcement Learning
Implementation of Generative Adversarial Imitation Learning (GAIL) on the OpenAI Gym CarRacing-v2 environment. The agent learns driving behavior purely from expert demonstrations without hand-crafted reward functions, using a discriminator to distinguish expert from agent trajectories.
Inverse RL · No explicit reward
SLAM + Object Detection in Complex Environments
Jan 2022 – May 2022 · ASU Research Assistant
AERO — Autonomous Explorer Bot
GPS-free autonomous exploration robot (ROS, Gazebo): Gmapping-SLAM for occupancy grid mapping, AMCL localization, ROS Navigation Stack with Dijkstra path planning, YOLOv4-Tiny object detection at 5 fps (CPU-only). AERO produced 75% AP, 90% tracking accuracy, 92% segmentation accuracy, 30% better depth estimation, and 25% compute reduction with Hybrid A*.
Hardware prototype · Real SLAM
Low-Cost EEG Cursor Control
Apr 2022 · Independent neural-interface demo
Cursor Controlling Using Brain-waves
Brain-computer interface prototype using a NeuroSky MindWave Mobile 1 headset to control cursor movement. The system establishes headset-to-PC communication, decodes attention, meditation, and eye-blink signals, then maps signal combinations into cursor actions for users with severe motor impairment.
Independent demo · Full PhD funding offer for neural-interface research
DP
Visual Deep Learning Model Builder
Visual Deep Learning Model Builder
May 2021 · Personal ML tooling project
Deep Playground
Graphical interface concept for building deep learning models: drag-and-drop layer creation, custom layers, graph-based dynamic path building, TensorFlow backend, one-click compilation, generated Python scripts, save/resume workflows, cloud export, and live training visualization.
Personal project · ML tooling
Master's Thesis
Multi-Robot Coordination in Unstructured Environments
Arizona State University · 2022 · Advised by Prof. Dr. Sangram Redkar
This thesis presents a comprehensive multi-robot coordination system integrating Unmanned Ground Vehicles (UGV) and Unmanned Aerial Vehicles (UAV) for search and rescue in unstructured environments. The system demonstrates autonomous navigation, real-time object detection, gesture-controlled drone swarms, and seamless UGV-UAV handoff for blocked-path scenarios.
State Estimation
Bayes filter, Kalman filter, Extended Kalman Filter for robot localization
SLAM & Navigation
ROS Gmapping occupancy grid, AMCL localization, Dijkstra + A* path planning
UAV AI Capabilities
MediaPipe Pose (33 3D landmarks) + KNN for body gesture control; 21 3D hand landmarks for gesture recognition
Object Detection
YOLOv4 on UGV and UAV for real-time human detection and object identification
Swarm Architecture
Decentralized multi-drone coordination with Mission Pad support; trajectory planning UI with collision-free paths
Hardware Prototype
Custom UGV with RPi 4B, Arduino Mega, RPLidar A1, L298N motor drivers, 12V DC motors + encoders
UGV-UAV Coordination
6-state decentralized state machine for seamless task handoff between ground and aerial robots
Thesis system overview
Multi-robot coordination workflow
UGV and UAV navigation stack
Gesture and perception pipeline
Search and rescue demo results
Research & Coursework
Academic work.
Additional research, coursework projects, and paper reviews from graduate and undergraduate studies.
MAE 547 · ASU · Jan–May 2022
Crowd-Aware Robot Navigation
Attention-based deep reinforcement learning (SARL / LM-SARL) for socially-compliant mobile robot navigation in dense crowds. Implemented self-attention pooling of human-human and human-robot interactions.
EEE 511 · ASU · Aug–Dec 2021
StackGAN-v2: Text-to-Image
Stacked Generative Adversarial Networks for high-resolution, photo-realistic image synthesis conditioned on text descriptions. Ran StackGAN-v2 on the bird dataset, validated sentence interpolation, inception score behavior, and failure cases/mode collapse.
SES 598 · ASU · Jan–May 2022
Local Path Planning for Self-Driving
Real-time path modification via cubic polynomial formulas; Voronoi cell-based collision avoidance algorithm for self-driving cars in complex obstacle environments.
Skills
Technical expertise.
Languages
Perception & State Estimation
Inference & Optimization
Frameworks & Tools
Also familiar with
MediaPipeGmapping SLAMDJI Tello SDKA*EKFAMCLAutowareLMDeployHuggingFaceEducation
Where I learned.
Aug 2021 – Dec 2022
M.S. Robotics and Autonomous Systems
Arizona State University
Tempe, Arizona
4.0 / 4.0
GPA
Thesis
"Multi-Robot Coordination in Unstructured Environments"
2015 – 2019
B.Tech Electronics & Communication Engineering
NIT Nagpur
Nagpur, India
Thesis
"GANs for Semi-Supervised Image Classification (F1: 0.95)"
When I'm not coding
DIY AI & Robotics
Tinkering after hours
3D Printing
From CAD to part
Travel
New places, new ideas
Cooking
Recipe iteration
Tara
My walking buddy
Contact
Let's build something.
Whether it's a new role, collaboration, or a project in autonomy, perception, or robotics — I'd love to hear from you.