Skip to content
View VedhSontha's full-sized avatar

Highlights

  • Pro

Block or report VedhSontha

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please donโ€™t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this userโ€™s behavior. Learn more about reporting abuse.

Report abuse
VedhSontha/README.md

๐Ÿงฌ About Me

B.Tech student at Manipal Institute of Technology (CCE, CGPA 7.93/10) โ€” building production-grade AI systems at the intersection of deep learning, computer vision, and edge hardware.

  • ๐Ÿ”ฌ 2 Provisional Patents filed (Low-Light Vision + Multi-Spectral Satellite Fusion)
  • ๐Ÿš€ Deployed real-time inference pipelines at 35 FPS on NVIDIA Jetson Nano via TensorRT INT8/FP16
  • ๐Ÿ›ฐ๏ธ Built satellite fusion networks achieving 79% mAP โ€” +18% over single-network baseline
  • ๐Ÿง  Mentoring 20+ undergraduates in Deep Learning and Computer Vision at Rugved Systems
  • ๐ŸŽฏ Targeting: AI/ML Research Internship (Computer Vision ยท Edge AI ยท Robotics)

โšก Tech Stack

Languages & Core

Python C++ Bash SQL

Deep Learning & CV

PyTorch TensorFlow Keras OpenCV Scikit-Learn YOLO

Edge & Deployment

TensorRT ONNX NVIDIA Jetson Docker

Robotics & Perception

ROS2 Open3D ArduPilot MediaPipe


๐Ÿ† Patent-Pending Research

๐Ÿ›ฐ๏ธ AMSSFN

All-Weather Multi-Spectral Satellite Fusion Network

PyTorch Multi-modal DL Patent Pending

  • Invented environment-conditioned cross-spectral attention โ€” atmospheric state (clear/cloudy/smoky/night) directly modulates SWIR-thermal-visible band interactions
  • Thermal super-resolution from 100m โ†’ 30m using learned Lagrange multipliers, preserving temperature accuracy within 0.2K
  • 79% mAP vs. 67% single-network baseline (+18% relative gain)

๐ŸŒ™ Intelligent Adaptive Low-Light Vision

Hardware-Adaptive Zero-DCE Edge System

PyTorch TensorRT NVIDIA Jetson Nano Patent Pending

  • Architected degradation-aware enhancement model superseding Zero-DCE; adaptive to high-variance noise and motion blur
  • Deployed at 35 FPS / ~15ms latency on Jetson Nano via TensorRT INT8/FP16 quantization
  • Dynamic hardware routing via gating mechanism along quality-vs-latency Pareto frontier

๐Ÿš€ Featured Projects

๐Ÿง  Brain Tumor Classification & Detection

CNNs YOLO PyTorch Paper Submission in Progress

  • 96% classification accuracy on MRI data via custom CNN architecture
  • Integrated YOLO bounding-box detection for precise tumor localization
  • Full ML pipeline: preprocessing โ†’ augmentation โ†’ training โ†’ Precision/Recall/F1/AUC evaluation

๐Ÿ”ง AstraSRE โ€” AI Anomaly Detection & Self-Healing

Isolation Forest Groq API Ollama LLM

  • Trained Isolation Forest on multi-signal observability (CPU%, latency, error rates); CRITICAL anomalies trigger auto-remediation achieving ~4s MTTR
  • LLM-powered incident reporting (Groq/Ollama) generates root-cause analysis from raw metric vectors; 100% report availability

๐Ÿš Autonomous Quadcopter โ€” Mono-Vision Navigation

ArduPilot Pixhawk ROS2 C++

  • GPS-denied navigation reducing flight path deviation by ~40% vs. open-loop baseline
  • Coordinate frame transformations + IMU sensor fusion for stable dynamics in unstructured environments
  • Designing decentralized multi-agent path planning for a swarm of 3 drones

๐Ÿค– VIO-Based Pathfinding & Mapping Robot

VINS-Mono RTAB-Map ROS2 A* C++

  • Engineered Visual Inertial Odometry (VIO) fusing monocular vision + high-frequency IMU for precise state estimation in GPS-denied zones
  • Real-time loop closure and occupancy grid mapping via RTAB-Map
  • A* path planning on live occupancy grid for dynamic obstacle avoidance

๐ŸŒ 3D Environment Reconstruction with RGB-D

Open3D ICP PCL Python

  • Multi-view reconstruction pipeline via ICP point cloud registration
  • Sub-centimeter alignment accuracy across 20+ viewpoints
  • Intrinsic/extrinsic camera calibration for precise depth-to-RGB alignment

๐Ÿ›ก๏ธ Cognitive Business Automation Suite

Streamlit Scikit-learn XGBoost SQLite

  • Customer churn prediction dashboard with Plotly risk gauge and dynamic retention actions
  • FlowbitAI: adaptive invoice parser with fuzzy self-correction and human-in-the-loop review
  • Rossmann store sales forecasting via pipeline-driven regression models

๐Ÿ’ผ Experience & Leadership

Role Organization Period
๐Ÿ›ธ Core Member & Technical Mentor โ€” AI Perception for Defense UAVs Rugved Systems Apr 2025 โ€“ Present
๐Ÿ“š Management Committee Member ACM Manipal Present

At Rugved Systems, leading AI/perception research for autonomous defense UAVs โ€” real-time threat detection using YOLO and ROS2-based deployment pipelines, while mentoring 20+ undergraduates in model training, evaluation, and simulation workflows.


๐Ÿ… Honors & Achievements

๐Ÿฅ‡ Achievement Details
๐Ÿ‡ฎ๐Ÿ‡ณ Smart India Hackathon (SIH) โ€” Round 2 Qualifier Developed Physics-Aware Multi-Modal Thermal Super-Resolution system; top national qualifier
๐Ÿค– E-Yantra Robotics Competition (IIT Bombay) โ€” Top 15 Recognized for AI-driven autonomous navigation system design

๐Ÿ“Š GitHub Stats



Pinned Loading

  1. Adaptive-HyperDCE-Edge Adaptive-HyperDCE-Edge Public

    A patent-pending, edge-optimized neural network for zero-reference low-light image enhancement. Features dynamic complexity-driven model routing and hardware-adaptive loss constraints.

    Jupyter Notebook

  2. binance-futures-trading-bot binance-futures-trading-bot Public

    Python

  3. Cognitive-Business-Automation Cognitive-Business-Automation Public

    A suite of machine learning models and intelligent automation tools designed to optimize business operations, featuring predictive sales forecasting, customer churn risk analysis, retrieval-based sโ€ฆ

    Jupyter Notebook

  4. drone-obstacle-detection drone-obstacle-detection Public

    Autonomous drone obstacle detection and collision avoidance system using a hybrid YOLO + DroNet fusion engine with depth-aware safety scoring.

    Python

  5. miscellaneous miscellaneous Public

    A consolidated repository of miscellaneous Python scripts, helper modules, and Jupyter Notebooks covering deep learning (CNNs, RNNs, Attention), computer vision (3D face reconstruction, RealSense lโ€ฆ

    Jupyter Notebook