Agentic AI & ROS Automation Engineer

AZARUDDIN SHAIK

AI Engineer Gen Engineer Automation AI Specialist Technical Staff Member

Software Engineer R&D Engineer ROS Developer Automotive • Agentic AI • Algorithm Development • Data Analytics *************@*****.*** +91-906******* Bengaluru, India linkedin.com/in/azaruddin-shaik0b3b0a207 PROFESSIONAL SUMMARY

AI Engineer and Software Engineer with 5 years of experience in the automotive domain: 2 years building LLM- powered agentic AI systems and 3 years in algorithm development and ROS 1/2-based systems for industrial automation and autonomous vehicle platforms. Built the SCN Coding Assistant for Mercedes-Benz — a multi-agent application (LangGraph, LangChain, Pydantic, Streamlit) that connects LLMs (Claude Sonnet/Opus, ChatGPT Terra, Kimi) to backend REST/Swagger APIs, automates SGDE/AED file generation, compares software releases, and answers coding-data queries using MCP, RAG and AWS Bedrock. Delivered 9 end-to-end projects, including 15+ ROS2 state machine states, Autoware-based Lanelet2 map generation across 5 real-world datasets, and embedded deployment on NVIDIA Orin. Achieved 40%+ improvement in turn-detection accuracy through sensor fusion and angular velocity filtering. Skilled across the full SDLC—from feasibility studies and architecture design to static analysis (Polyspace), unit testing (VectorCast), and production deployment via Git/Linux. CORE TECHNICAL SKILLS

Languages Python, C++

Agentic AI & LLM

Frameworks

LangGraph (state-machine agents), LangChain, Pydantic (structured outputs), Prompt Engineering, RAG, MCP (Model Context Protocol), Tool/Function Calling, GenAI LLMs Claude Sonnet, Claude Opus, ChatGPT (Terra), Kimi, AWS Bedrock, OpenAI & Google GenAI SDKs

Machine Learning Scikit-learn, Logistic Regression (intent classification), Joblib, regex-based entity extraction Backend & Integration REST / Swagger APIs, OAuth token flows, Requests, Response Caching, WebSocket, AWS

(Bedrock, boto3, EC2)

UI & Data Tooling Streamlit, Pandas, OpenPyXL, XlsxWriter, NumPy, Matplotlib, Open3D, Python-dotenv Robotics Frameworks ROS1 (Noetic), ROS2 (Humble/Jazzy), RCLCPP, RCLPY Autonomous Systems Autoware, Nav2, SLAM Toolbox, Cartographer, D* Lite, LIO-SAM Mapping & Sensors Lanelet2, LiDAR, Camera, IMU, GPS, ROSBAG, RViz, Gazebo Algorithms Motion Matrix, RANSAC, Cubic Spline Interpolation, State Machines (Yasmin) Tools & Platforms Git, Linux, AWS EC2, NVIDIA Orin, WebSocket, Polyspace, VectorCast, Vector Map Builder, JOSM

PROFESSIONAL EXPERIENCE

AI Engineer / Gen Engineer / Automation AI Specialist / Technical Staff Member — KPIT Nov 2025 – Present Designing and building agentic AI applications (SCN Coding Assistant) for client Mercedes-Benz using LangGraph, LangChain and multiple LLMs, and working on ROS-ECU bring-up for the Mercedes-Benz ATG autonomous car platform.

Software Engineer / R&D Engineer / ROS Developer — Haystack Robotics Aug 2025 – Nov 2025 Developed an autonomous medical disinfection robot for hospital environments using ROS2. Software Engineer / R&D Engineer / ROS Developer — Wipro Technologies Pvt. Ltd. Dec 2021 – Aug 2025 Led development of ROS2 industrial automation systems and Autoware-based HD map generation pipelines for Oshkosh Corporation, delivering 6 projects over 3.7 years across 5 real-world autonomous driving datasets. PROJECT EXPERIENCE

Project 9: SCN Coding Assistant — Agentic AI Application Mercedes-Benz / KPIT Nov 2025 – Present Tech Stack: Python, LangGraph, LangChain, Pydantic, Streamlit, RAG, MCP, AWS Bedrock, REST/Swagger APIs, Scikit-learn (Logistic Regression), Pandas, Git, GenAI

• Developed an agentic AI application that connects to backend RESTful (Swagger) APIs through a ChatGPT/Claude-style chat interface, built with LangGraph and LangChain and delivered through a Streamlit UI.

• Built and evaluated the agents using multiple LLMs — Claude Sonnet, Claude Opus, ChatGPT (Terra) and Kimi

— and applied AWS Bedrock concepts for LLM access.

• Automated SGDE and AED file downloads for specific carlines using UUID-based API requests, CodeMan links, and CodeMan version + ECU variant (multi-carline), and built a bulk flow to generate SGDE/AED files for bulk vehicles.

• Trained a Logistic Regression model (Scikit-learn) to classify user intent and route requests to the correct SGDE/AED download flow.

• Implemented a Coding Data Agent (LangGraph + Pydantic) that answers natural-language queries on coding data, and displays the carline results on the UI.

• Developed a Comparison Agent that compares two released software versions, diffs their carline data, and presents the differences on the UI.

• Integrated an MCP server to access artifact pages and retrieve release documents, and applied RAG to fetch relevant information and display it on the UI.

Project 8: ROS-ECU Bring-Up (Minerva ATGB Car) Mercedes-Benz / KPIT Nov 2025 – Present Tech Stack: C++, ROS1, RCLCPP, WebSocket, IMU, Vehicle Signals, NVIDIA Orin, GenAI

• Implemented 4 dummy vehicle signal publishers (wheel angle, acceleration, deceleration, gear position) as ROS1 nodes using custom message definitions.

• Developed WebSocket-based middleware to bridge ROS components with non-ROS IDC (ECU) systems.

• Flashed and deployed non-ROS applications on NVIDIA Orin; validated end-to-end signal transmission across 4 signal types on the ATGB vehicle.

• Implemented IMU and wheel data ingestion from IDC (ECU) into the ROS environment; recorded data streams using ROSBAG.

• Gained hands-on experience with hardware interfacing and network configuration on Linux and NVIDIA Orin platforms.

Project 7: Autonomous Medical Disinfection Robot Haystack Robotics Aug – Nov 2025 Tech Stack: Python, ROS2 Jazzy, RCLPY, SLAM Toolbox, Nav2, D* Lite, LiDAR, Depth Camera, Gazebo, RViz

• Designed end-to-end system architecture for an autonomous hospital disinfection robot capable of real-time navigation and obstacle avoidance.

• Developed a D* Lite global path planning proof-of-concept enabling dynamic re-routing in changing environments.

• Integrated SLAM Toolbox with Nav2 stack for simultaneous real-time mapping and autonomous navigation.

• Implemented behavior trees for structured decision-making during disinfection task execution. Project 6: Autoware Lanelet2 Map Generation — Phase 3 Oshkosh Corporation / Wipro Dec 2023 – Aug 2025 Tech Stack: Python, NumPy, Pandas, Matplotlib, RANSAC (skimage), AWS EC2, Git, GenAI

• Developed an intersection detection module capable of identifying turns across 5 diverse dataset geometries, covering both right-angle and S-curved road types.

• Eliminated IMU noise interference by replacing raw IMU-based turn detection with angular velocity filtering, achieving 40%+ improvement in turn-detection accuracy across all tested datasets.

• Designed an adaptive windowing technique for dynamic segmentation of S-curved roads using GPS data and a 10 m distance threshold with residual analysis.

• Implemented midpoint correction logic to resolve geometric misalignments between adjacent road segments.

• Explored and evaluated camera-to-LiDAR calibration workflows from Autoware documentation. Project 5: Autoware Lanelet2 Map Generation — Phase 2 Oshkosh Corporation / Wipro Dec 2023 – Aug 2025 Tech Stack: Python, NumPy, Pandas, Matplotlib, RANSAC, Cubic Spline, AWS EC2, Git

• Implemented turn detection using IMU data for both right-angle and curved road segments.

• Applied RANSAC algorithm for robust road geometry line fitting on segmented path data.

• Developed cubic spline interpolation with equilateral triangle logic to generate smooth circular arcs at road transitions.

• Determined turn direction (left/right) via cross-product mathematics applied to road segment vectors. Project 4: Autoware Lanelet2 Map Generation — Phase 1 Oshkosh Corporation / Wipro Dec 2023 – Aug 2025 Tech Stack: Python, NumPy, Pandas, Matplotlib, ROS2 Nav Stack, Autoware, Cartographer, LIO-SAM, AWS EC2

• Designed and implemented 3 core pipeline modules — Data Unpack, Data Synchronization, and Motion Matrix Algorithm — enabling OSM-to-PGM conversion across all 5 Oshkosh datasets.

• Validated pipeline output by integrating PGM maps with the ROS2 Navigation Stack and Lanelet2 maps with Autoware for Turtlebot and vehicle simulations.

• Researched SLAM techniques (Cartographer, LIO-SAM, LIO-RF) and lane detection models (UltraFast-Lane- Detection) for pipeline integration.

• Authored Lanelet2 maps manually using Vector Map Builder and JOSM; documented README guides for map loading workflows.

Project 3: Autonomy Arm Stack — Cart Pickup System Phase 2 Oshkosh Corporation / Wipro Jan 2022 – Nov 2023

Tech Stack: C++, ROS2, RCLCPP, Yasmin State Machine, Polyspace, VectorCast, Git, RViz

• Designed and implemented the full Autonomy Arm Stack State Machine in C++ using the Yasmin Library, covering 7 operational states (System Ready System Enabled Cart Detected Grabbing Dumping Placing Back Default) with 100% unit test coverage via VectorCast.

• Built ROS2 communication architecture with custom messages, CRC checksums, and message counters for reliable inter-node data transmission.

• Developed auxiliary modules: Human Detector (ROS2 node) and Health Safety Monitor (C++ component), integrated into the core state machine.

• Conducted static code analysis (Polyspace) and unit testing (VectorCast) across 7 states; created comprehensive test scenarios and flowcharts for each state, achieving zero critical defects at integration. Project 2: Cart Seeker & Robot Arm State Machine — Phase 1 Oshkosh Corporation / Wipro Jan 2022 – Nov 2023 Tech Stack: C++, ROS2, RCLCPP, Yasmin State Machine, Polyspace, VectorCast, Git, RViz

• Implemented 7 Cart Seeker and Robot Arm State Machine states (System Ready, System Enabled, Cart Detected, Grabbing, Dumping, Placing Back, Default) using Yasmin Library in C++, delivering the first functional ROS2 automation PoC for the client.

• Established publisher-subscriber communication for Arm Sensor and Cart Detector with CRC checksums and message counters for data integrity.

• Presented state machine proof-of-concept to stakeholders and SMEs; participated in full SDLC including design reviews, code reviews, and version control.

Project 1: Standard State Machine — ROS1 Proof of Concept Oshkosh Corporation / Wipro Jan 2022 – Nov 2023 Tech Stack: Python, ROS1, RCLPY, Tkinter GUI, RViz, Git

• Developed 4 ROS1 publisher nodes (Random Numbers, Sine Wave, Sawtooth Waveform, Deadman Switch) and a ROSBAG recorder capturing all outputs for replay and validation.

• Built a threshold-based subscriber that aggregated inputs from multiple nodes and implemented a state machine transitioning from Initial to System Ready state.

• Created an RViz image visualization pipeline and a Tkinter GUI for real-time system monitoring.

• Demonstrated proof-of-concept to stakeholders after thorough testing with SMAC Viewer. EDUCATION

B.Tech in Computer Science and Engineering Malla Reddy College of Engineering and Technology, Hyderabad 2020

Contact this candidate