ANKIT SATYENDRA KUMAR VERMA

Detail-oriented AI/ML Engineer with expertise in Python programming and machine learning frameworks. Proficient in data analysis and developing innovative solutions to complex problems. Strong communication and teamwork skills, demonstrated through successful project collaborations.

• IELTS: Overall band score 7.0 & CERF Level C1
• 100 Days of Code: The Complete Python Pro Bootcamp (Udemy, Issued July 2023)
• Machine Learning A-Z: AI, Python & R + ChatGPT Prize [2024] (Udemy, Issued Dec 2023)
• Must Learn Programming for every Freshers - Java & Python (Udemy, Issued July 2018)
• Received Accenture Spotlight Award for collaboration with client.

Drone Camera to Map Identification

• Designed and deployed machine learning models (LSTM, deep neural networks) in Python and PyTorch to solve time-series regression problems using high-frequency sensor data.
• Built end-to-end data science pipelines, including data preprocessing, feature engineering, sequence modelling, and train/validation/test evaluation on large real-world datasets.
• Improved model performance through hyperparameter tuning and benchmarking, achieving ~75% reduction in prediction error and ~97% reduction in drift-related error over baseline approaches.
• Validated models using standard ML regression metrics (MSE, RMSE, error analysis), demonstrating robust generalization, low validation loss, and stable convergence.
• Frameworks - Python, PyTorch, NumPy, SciPy, Pandas, TensorFlow, scikit-learn, CUDA

Emergency Landing Decision Support for Urban eVTOLs

• Developed an AI-driven decision-support system using computer vision and reinforcement learning to identify safe landing zones and guide emergency navigation in urban environments.
• Trained computer vision models (VGG16, YOLOv8) for classification and object detection, achieving 97% accuracy and 0.72 precision / 0.61 recall.
• Built and evaluated reinforcement learning models (Actor–Critic, DQN), reducing average task completion time by ~30% through data-driven optimization.
• Analyzed real physiological data (heart rate) to quantify user impact, achieving ~40% reduction in cognitive load via optimized AI-generated guidance.

Frameworks used – Python, PyTorch, ROS2, Gazebo, YOLOv8, VGG16, OpenCV, CUDA