ADDITYA AKASH MISHRA

AI Engineer

Professional Experience


Laboratory for Electro-Optics Systems (LEOS) - ISRO
Project Trainee
Aug 2025 - Present
  • Designing algorithms to generate mirror surface profiles from deflection angle data for aspheric mirror optical metrology
  • Applying statistical and ML methods for real-time parameter estimation to enhance algorithm accuracy
  • Automating data processing workflows to eliminate human bias and ensure consistent, repeatable results
GE Appliances
Digital Technology Intern
Jan 2024 - July 2024
  • Resolved critical bugs and enhanced authentication for smart home device applications, ensuring security and reliability
  • Developed automated workflows for DynamoDB queries, improving operational efficiency by 95%
  • Delivered multiple tasks ahead of schedule through effective time management and automation
  • Gained hands-on experience with AWS cloud services for scalable application enhancements
Manish & Raj Associates
Freelance Web Developer
Aug 2023 - Sep 2023
  • Developed Flask-based web application with SQLAlchemy for streamlined database management
  • Created customized billing system automating client billing processes and improving efficiency
  • Built comprehensive customer data management module enhancing organization and accessibility
  • Delivered user-friendly solution reducing processing time by 99% through seamless integration

Key Projects


Vision Navigation: Modular VSLAM
June 2025
  • Built a modular Visual SLAM pipeline implementing feature detection/matching, Essential/Fundamental matrix pose estimation, and local mapping components for state estimation and re-localization on KITTI odometry sequences.
  • Implemented Bag-of-Visual-Words for loop closure using KMeans vocabularies with cosine-similarity retrieval and geometric check, significantly reducing drift for robust GPS-denied navigation.
  • Desinged a realt-time Streamlit dashboard for inspecting keypoints, descriptor matches, BoVW histograms, and animated 3D trajectories, enabling fast debugging and evaluation of SLAM performance.
  • Developed a high-performance path for a VSLAM pipeline by exporting a learned feature-extraction module to ONNX and compiling it with TensorRT, achieving ≥ 30 Hz inference in desktop simulation targeting NVIDIA Jetson edge deployment.
Internet Researcher: Multi-Agent System
Oct 2025
  • Built 4-agent AI system for web search, documentation, validation, and CSV reporting of research papers
  • Designed reusable autonomous agents using agentic AI frameworks to automate research workflows