CV
Sarthak Chittawar
Undergraduate Researcher at IIIT Hyderabad
Summary
Computer Science and Engineering student at IIIT Hyderabad working on mobile robotics, computer vision, NLP, and machine learning systems. Undergraduate researcher at the Robotics Research Centre advised by Prof. K. Madhava Krishna.
Education
- Computer Science and EngineeringJuly 2026International Institute of Information Technology, Hyderabad
Work Experience
- AI Research InternMay 2025 - Aug. 2025IBM ResearchNLP, computer vision, document intelligence, and hyperspectral satellite modeling in Bengaluru, Karnataka.
- Engineered an end-to-end pipeline for extracting unstructured tabular data from PDFs into accurate CSVs, improving significantly on previous work.
- Developed QA and summary models using RAG-based few-shot prompting on an SQL backbone for user interaction.
- Worked on methane emission rate prediction from hyperspectral satellite data and local wind speeds by fine-tuning the Prithvi-V1-100 model.
- Received a pre-placement offer after completing the internship.
- Undergraduate ResearcherMay 2023 - PresentRobotics Research Centre, IIIT HyderabadMobile robotics and computer vision research advised by Prof. K. Madhava Krishna.
- Worked on MASt3R-guided topological navigation and pixel-relative representations.
- Published research on topological navigation at IEEE ICRA 2026 and IEEE/RSJ IROS 2026.
- Machine Learning InternJune 2024 - Aug. 2024Stealth StartupComputer vision and PyTorch internship in Hyderabad, Telangana.
- Developed a YOLO-based framework for detecting defects in electric power lines.
- Built model architecture components from scratch and managed data preprocessing and training.
Skills
Languages
- Unix Bash
- C/C++
- Python
- Java
- HTML/CSS
- SQL
- x86 Assembly
- JavaScript
- GLSL Shading Language
Technologies
- Linux
- Git
- MERN Stack
- Flask
- Docker
- THREE.js
- D3.js
- OpenGL
- PyTorch
- TensorFlow
- Weights & Biases
Publications
- PixelLoop: Shortcut Topological Navigation with Pixel-Level Loops2026IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)Shortcut topological navigation using pixel-level loop closures.
- MASt3R-Nav: WayPixel Navigation in Relative 3D Maps2026IEEE International Conference on Robotics and Automation (ICRA)WayPixel navigation in relative 3D maps for geometrically grounded robot navigation.
- Towards Global Localization using Multi-Modal Object-Instance Re-Identification2025Advances in Robotics (AIR) - OralMultimodal object-instance re-identification using RGB and depth for robust localization.
Projects
- MedminiOct. 2023NLP, LLM, React.js, Flask, DockerLightweight LLM-based medical data answering system for edge devices.
- Designed for less than 3 GB RAM, 0 VRAM, and sub-3-second inference.
- Deployed through the Android Expo Go app.
- Awarded 2nd place by Qualcomm at Megathon 2023.
- Image Style Transfer using TransformersApr. 2024Computer Vision, NLPEnhanced a CVPR 2022 transformer-based style transfer architecture with content-aware positional encoding.
- Integrated CLIP for text-based image retrieval.
- Used optical flow for video style transfer.
- Fast Inference using Speculative DecodingNov. 2024NLP, Machine Learning, PyTorchImplemented speculative decoding with knowledge distillation to improve inference times for translation and generation tasks.
- Improved on the baseline with online speculative decoding between draft and target models.
- A Study in Mechanistic Interpretability - GrokkingApr. 2024NLP, LLM, AI SafetyStudied mechanistic interpretability and AI safety through grokking experiments on a toy language model.
Achievements
- IndiaAI fellowship for graduate students2025IndiaAI
- 2nd place at Megathon 20232023QualcommAwarded for Medmini at Hyderabad's largest student hackathon.
- All India Rank 44 in IIIT-UGEE2021IIIT-UGEERanked among 9.4 lakh students.
- All India Rank 3887 in JEE Main2021JEE Main
Interests
- Research InterestsMobile Robotics, Computer Vision, Visual Navigation, Topological Mapping, NLP, Foundation Models, AI Safety