Atharv Gaur
Email: atharvgaur591@gmail.com | Phone: +91 89303 09419
|
GitHub |
LinkedIn
Explore My Work:
Projects |
Coursework |
Achievements
|
Research-Oriented Computer Science Student
I am an undergraduate in Computer Science Engineering at Birla Institute of Technology and Science (BITS), Pilani,
with a strong passion for Computer Vision, Machine Learning, and Robotics research. My academic journey is driven by a commitment to experimental design,
reproducibility, and meaningful contributions to state-of-the-art research.
Current Research & Work:
I am currently engaged in multiple research initiatives:
- Transformer-based Sensor Data Analytics with Prof. Pradeep Kundu at IIT Indore, developing real-time monitoring systems for industrial IoT applications.
- Neuro-Symbolic AI with the POSTMAN API Lab at BITS Pilani, working on zero-shot adaptation to non-stationary safety constraints in multi-agent reinforcement learning.
- Autonomous Robotics with CRISS Robotics at BITS Pilani, developing autonomous navigation and control systems for the ERC competition.
- Embedded Systems as a developer with IEEE BITS Pilani, designing autonomous maze-solving micromouse robots.
Technical Expertise:
My technical foundation spans deep learning architectures (CNNs, Transformers, LSTMs), computer vision applications,
robotics (ROS 2, SLAM, Sensor Fusion), and embedded systems programming. I work extensively with PyTorch, TensorFlow,
and specialized libraries for computer vision and robotics.
Notable Achievement:
I was selected as a Finalist (Top 0.3%) in the LAM Research Challenge 2025 out of ~8,000 registrations
for my Automated Lunar Traverse Planner, which integrated YOLO-based crater detection, solar-shadow-aware navigation,
and hierarchical path planning.
I am always open to collaborations, research opportunities, and meaningful discussions. Please feel free to reach out!
Other:
Achievements & Awards |
Interests & Activities
|
| [2025] |
LAM Research Challenge 2025 Finalist — Selected in top 0.3% (24 out of ~8,000) for Automated Lunar Traverse Planner project.
|
| [Jun 2025] |
Started Transformer-based Research Internship with Prof. Pradeep Kundu at IIT Indore on sensor data analytics for industrial IoT. |
| [Apr 2025] |
Joined CRISS Robotics team as Software Subsystem Developer, contributing to autonomous navigation systems. |
| [2025] |
Initiated research with POSTMAN API Lab on neuro-symbolic AI and safety-constrained multi-agent reinforcement learning. |
| [Sep 2024] |
Started as Embedded Systems Developer with IEEE BITS Pilani, designing micromouse robots and embedded systems. |
|
Deepfake Detection using CNN-LSTM
Technologies: PyTorch, OpenCV, ResNet-18, Bi-LSTM
Achievement: 87.69% accuracy, AUC-ROC ≈ 0.90
Built a hybrid deep learning model combining ResNet-18 CNN for spatial feature extraction and Bidirectional LSTM for temporal pattern analysis
across 16-frame video sequences. Implemented comprehensive face preprocessing with landmark alignment and data augmentation, achieving robust
balanced generalization with balanced F1 scores on both real and fake classes.
|
|
Automated Lunar Traverse Planner
Technologies: Python, NumPy, Matplotlib, YOLO crater detection
Achievement: LAM Challenge 2025 Finalist (Top 0.3%)
Developed an intelligent path planning system for lunar rover missions integrating YOLO-based crater detection with solar illumination-aware navigation.
Implemented solar shadow simulation using physics-based shadow polygons and designed hierarchical three-tier navigation with science-optimal waypoint
selection and automated regolith sampling site identification.
|
|
BDH Narrative Consistency Evaluator
Technologies: PyTorch, Transformers, Mechanistic Interpretability
Built inference and interpretability pipeline on Baby Dragon Hatchling (BDH) transformer architecture for narrative contradiction detection.
Engineered structured training data through synthetic contradiction generation and implemented mechanistic interpretability probes for high-resolution
attention visualization across transformer layers and heads.
|
|
Image Segmentation using U-Net
Technologies: PyTorch, Albumentations
Implemented complete semantic segmentation pipeline with U-Net architecture, custom DataLoader with train/validation splits, and real-time augmentations.
Applied Dice + BCE hybrid loss function and integrated GPU acceleration with reproducible seeding for scalable experimentation.
|
|
Micromouse Maze Solver
Technologies: C++, Teensy 4.1, VL53L0X, IMU — IEEE BITS Pilani
Designed autonomous micromouse robot with embedded systems integration using Teensy 4.1 MCU, VL53L0X ToF sensors, and IMU for precise navigation.
Implemented flood fill algorithm, PID control for lane-centering, and shortest-path logic for real-time maze solving.
|
Research Intern
Jun 2025 - Present
IIT Indore
Indore, India (Remote)
|
Working with Prof. Pradeep Kundu on Transformer-based architecture for real-time monitoring and classification of sequential sensor data in industrial settings.
Designing multi-head attention modules tailored to non-stationary time-series from industrial IoT sensors.
|
Research Member
2025 - Present
BITS Pilani
POSTMAN API Lab
|
Working on Zero-Shot Adaptation to Non-Stationary Safety Constraints in Multi-Agent Reinforcement Learning via Neuro-Symbolic Shielding.
Developing formal symbolic safety shields that integrate with MARL policies to enforce dynamic constraint satisfaction without retraining.
|
Software Subsystem Developer
Apr 2025 - Present
BITS Pilani
CRISS Robotics
|
Developing autonomous control systems and ROS2 communication modules for the ERC competition.
Built autonomous airlock control system with 10-state FSM logic and sensor fusion, and currently implementing autonomous navigation using ArUco markers and Nav2.
|
Embedded Systems Developer
Sep 2024 - Present
BITS Pilani
IEEE Technical Division
|
Designing and implementing autonomous micromouse robot with Teensy 4.1 MCU, VL53L0X sensors, and IMU.
Implementing flood-fill maze-solving algorithms with PID control and Hamming Code-based error correction systems.
|
|
Programming Languages: Python, C++, C, Java
Deep Learning & AI: CNNs, Transformers, LSTMs, GRUs, ResNets, U-Nets, Neural Style Transfer, Reinforcement Learning, GANs, RAG, Vision-Language Models, Mechanistic Interpretability
Frameworks & Libraries: PyTorch, TensorFlow, HuggingFace Transformers, Scikit-Learn, XGBoost, OpenCV, Albumentations, MTCNN
Data Science: NumPy, Pandas, Matplotlib, Seaborn, FFT for Time-Series Analysis
Robotics & Embedded: ROS 2 (Jazzy), SLAM, Sensor Fusion, Gazebo (Harmonic), RViz, Arduino IDE, ESP32, Teensy 4.1
|
|