class KaranSharma:
    def __init__(self):
        self.role        = "Robotics & Software Systems Engineer"
        self.education   = "MEng Software Systems @ UTS Sydney (2026)"
        self.experience  = "3+ years in autonomous systems & embedded software"
        self.currently   = "Building real-time robotic systems in Sydney 🇦🇺"

    def tech_stack(self):
        return {
            "robotics"   : ["ROS2", "SLAM", "LiDAR", "ArUco Markers", "Path Planning"],
            "control"    : ["PID", "LQR", "Adaptive Control", "Inverted Pendulum"],
            "embedded"   : ["ESP32", "ESP-IDF", "LVGL", "Jetson Nano", "IMU/Gyro"],
            "languages"  : ["Python", "C++", "Rust"],
            "platforms"  : ["Ubuntu/Linux", "ROS Foxy", "Git", "AWS"],
            "ai_ml"      : ["Deep Learning", "Multispectral Imaging", "Predictive Modelling"],
        }

    def passion(self):
        return "Bridging hardware & software to build robots that work in the real world 🚀"

A full ROS2 package that adds a natural-language voice interface and LLM behaviour engine to a TurtleBot3. Bridges offline speech recognition (faster-whisper), a locally-hosted LLM (Ollama/qwen2.5-coder), and Nav2 navigation into a single coherent pipeline.

Key features:

• 🎙️ Wake-word gated voice control ("Hey Pepper") with offline STT — no internet required at runtime
• 🤖 LLM intent classification with rule-based fast path (eliminates ~500ms latency for safety-critical commands)
• 🗺️ Tour-guide behaviour — records routes via teleop, replays with waypoint narration via TTS
• 🧠 Behaviour tree orchestrator for navigation, recording, pause/resume, and localisation recovery
• 😊 Emotion/gesture engine publishing to Pepper robot gesture controller in real-time
• 🖥️ PyQt5 TSP-optimal waypoint GUI for tour planning

Stack: ROS2 Humble · Python 3.10 · faster-whisper · Ollama · Nav2 · PyQt5 · Behaviour Trees

📄 Full Documentation · 🔗 Repository

Designed and commissioned a two-wheeled robot using DDSM115 servomotors and benchmarked 6 control architectures: PID, PI, PD, PDD², LQR, and Adaptive Control — analysing stability margins, transient response, and disturbance rejection under load.

Full-stack autonomous navigation system on ROS2 (Foxy) with LiDAR SLAM for 3D point cloud generation and occupancy mapping. Deployed on Jetson Nano for real-time edge inference across unstructured environments.

Programmed flight control algorithms for sub-centimetre positional accuracy using ArUco marker detection and IR-based localisation for autonomous takeoff, hover, and landing.

Built a real-time telemetry analysis platform in Python for in-flight sensor monitoring — cutting anomaly response time by 50% with 100% command success rate across all mission-critical transmissions.

Scalable device provisioning framework with LVGL-based GUI firmware integrating real-time multi-sensor acquisition (gyro, barometric pressure, voltage, current) with live telemetry visualisation.

Multispectral UAV imagery pipeline for precision agriculture, producing a deep learning predictive model with measurable accuracy improvements. Research paper submitted for publication.

At Plaksha: Investigated 1D humanoid balance dynamics with Prof. Andy Ruina · Built autonomous airship gondolas with Prof. Rajkumar Pant (IITB) · Taught "Intelligent Machines" and "Control, Autonomy & Planning"

At National Ilan University: LiDAR SLAM on mobile robots with ROS2 · Blockchain-backed IoT edge devices · DJI Tello precision UAV control

• 📚 MEng Software Systems @ University of Technology Sydney (Expected Dec 2026)
• 🔭 Exploring maritime & field autonomous systems
• 🌊 Interested in ROS2-based maritime robots, perception systems, and real-world deployment
• 📬 Open to internships, research roles, and robotics projects in Sydney

"The best robot is the one that works in the real world, not just in simulation."

⭐ If you find my work interesting, let's connect!