devarshi

DEVARSHI AGGARWAL

Bengaluru +91 9792910973 devershigpt6@gmail.com linkedin/devarshi-aggarwal/

Preview of Devarshi Aggarwal CV PDF PDF

SUMMARY

A dynamic and innovative Deep Learning Engineer with hands-on experience in computer vision, machine learning, and artificial intelligence. Have demonstrated expertise in developing and deploying cutting-edge solutions for diverse applications such as footfall counting, licence plate recognition, and employee punctuality scoring. Proven ability to leverage advanced tools and technologies like PyTorch, OpenCV, and Docker enables the team to deliver scalable and efficient real-time solutions. With excellent problem-solving abilities, consistently deliver high-quality projects within deadline constraints. Skilled in extracting actionable insights from complex datasets, driving impactful decision-making. Adept at collaboration and communication, with a passion for continuous learning and staying updated with the latest advancements in the field.

EXPERIENCE

Samsung Electro Mechanics, India

Deep Learning Engineer December 2024 - Present

  • Worked on dual-arm titan robot Isaac simulator reinforcement learning and imitation learning training. Improved task success rate from 0% to 99% after running 100+ RL and IL experiments.
  • Worked on a driver management system with features including gaze direction and sleepiness detection. Trained face detection, identification, and facial landmark detection models with 90%+ accuracy.

Jio, India

Deep Learning Engineer April 2022 - December 2024

  • Worked on novel architecture for an optimized deployment of people footfall counting, heatmap and flowmap.
  • Improved detection in JioANPR system from 90% to 95%. Improved recognition accuracy from 83% to 94%. Overall accuracy of pipeline post tracking is at 99%. Developed reproducible retraining pipeline.
  • Leveraged NVIDIA-Deepstream for efficient and scalable real-time solution. In JioFootfall, increased camera density per GPU from 8 cameras to 108 cameras.
  • Machine learning solution for punctuality score service of employees.
  • For face detection and person access control, leveraged vector database Milvus for real-time feature vector search with very low lookup latency.

Agro Informatics Lab, IIT Bombay, India

Computer Vision Intern Aug 2021 - March 2022

  • Extracted differentiating features from images of various genotypes of rice.
  • Used machine learning techniques to identify the best species for inter-species breeding.
  • Made tools and scripts for data labelling and data augmentations which greatly improved manual labelling time by 75%.

Freelancer

Freelancer July 2020-July 2021

  • Worked for various clients on small and medium-sized projects using python and deep learning skills.
  • Developed a deep reinforcement learning model for trading on the metatrader tool, for an independent client.

Department of Computer Science, IIT Bombay

Machine Learning Intern July 2019- May 2020

  • Mentor: Pankaj Singh, Project Research Associate, IIT Bombay
  • Designed and developed an annotation tool Form Labeller using Python and Tkinter for efficient data labelling for the use case.
  • Created a Form Schema detection three-step pipeline, with box detection, OCR and then schema detection which could extract and tabulate information from hand-written or printed forms in <1s.
  • Obfuscated, binarized and licensed two projects for clients using pyarmor and pyinstaller.

Publications

Leaf Count Aided Novel Framework for Rice (Oryza sativa L.) Genotypes Discrimination in Phenomics: Leveraging Computer Vision and Deep Learning Applications
On 3-Degree 4-Chordal Graphs

Found polynomial time algorithms for classical NPC problems like vertex cover, dominating set, feedback vertex set for the graph subclass - maximum degree 4 max chordality 3

EDUCATION

IIITDM-Kancheepuram – Chennai, India

Bachelor of Engineering, Computer Science July 2015- June 2019

  • Cumulative GPA: 8.18/10.0

  • Relevant coursework: Artificial Intelligence Search Methods for Problem Solving, Pattern Recognition, Digital Image Processing, Joy of Computing using Python, Operating Systems, Probability Theory, Linear Algebra, Differential Equations, Calculus, Design and Analysis of Algorithms, Database Systems

Sunbeam School Lahartara – Varanasi, India

Secondary School July 2012- June 2014

  • Cumulative GPA: 9.3/10.0

SKILLS

Programming Languages: Python(Expert), SQL(Expert), C(Intermediate), Shell(Intermediate) , Rust (Intermediate) Tools: PyTorch, Numpy, Pandas, Sklearn, OpenCV, Kubernetes, Docker, Milvus, Deepstream, Ultralytics, Locust, Streamlit
Languages: Fluent in English, Hindi

OTHER ACHIEVEMENTS

  • Secured AIR-5320 (GEN) in IIT JEE-Advanced 2015