CV

Click here to get Resume

Education

• M.S. in Computer Science, University of Southern California, 2024-2026
• B.S. in Artificial Intelligence and Data Science, Vishwakarma Institute of Information Technology, 2020-2024 (now VIT )

Work experience

• Research Assistant: VISHWAKARMA INSTITUTE OF INFORMATION TECHNOLOGY
  • Advisor: Parikshit Mahalle [Ph.D.],
  • Conducting advanced research in the intersection of Generative AI and Computer Vision for the automation of Cranial Implant Design to shorten the time and cost overhead of designing Patient-Specific Implants
  • Calibrated a novel dataset of 930+ augmented samples as a baseline to test AI models
  • Tested the reconstruction abilities of a Generative Adversarial Net [GAN] and a Sparse Tensor based CNN using Nvidia A100 GPU. The GAN increased the dice score to 92% while the Tensors enabled CNN induce sparsity to accelerate the computations and enhance robustness by reducing GPU-RAM usage from 40 GB to 6 GB
• UX Architect Intern: HTS TECH SOLUTION
  • Designed the wireframe-enabled architecture and data pipeline maps for end-to-end modules like lead management, inventory management, and ledger management for a large Real Estate development firm
  • Integrated a Key Performance index Layer to monitor progress in various domains for the client with an improvement in System Usability Score by 13% and Task Success Rate by 32%

Projects

• NEET-QA
  • Accepted at ICAAIML,
  • Implemented a distinctive application of RAG-based Question Answering System in the NEET (Indian Medical Entrance) exam using Langchain for Gemini-Pro, GPT-3.5-turbo-0125 LLMs with custom retrieval methods
  • Boosted answer accuracy by 12.5% on average against simple LLMs in Physics, Chemistry, and Biology MCQ answering, with the highest increment of 20.1% in the Biology domain. Work includes analysis of the generated answers and provides crucial checkpoints in advancing the RAG systems
• VOLUMETRIC IMAGES OF PATIENT-SPECIFIC IMPLANTS FOR CRANIAL DEFECTS
  • Data Set Published at Mendeley, Data Commons,
  • Published a dataset of 961 volumetric defect skull & implant pair samples derived from 29 base samples using Mesh-to-Volume Grid transformations, morphology, and augmentation by Affine, Non-Rigid Registration to facilitate the development and evaluation of AI models for auto-generating implants, perform clinical simulations to assess surgical complexities, and establish guidelines for efficient implant design, the proposed data can be utilized

Skills

• Languages:
  • Python, C++, C, C#,
• Framework:
  • Machine / Deep Learning
    • PyTorch, Tensorflow, Keras, Yolo, Langchain
  • Development:
    • FlaskAPI, React.js, Streamlit
• Libraries:
  • NumPy, PandasSKLearn, Matplotlib, Seaborn, OpenCV, ANTs, FAISS, Sharp
• Databased:
  • MongoDB, MySQL, Chroma
• Tools:
  • Git / Github, Docker, Unity 3D, Figma, Adobe Illustrator

Publications

NEET-QA

Dhiren Oswal, Soham Tolwala, Pushpak Suryawanshi, Dhairya Thakkar, Swapnil Shinde (2024). " NEET-QA: Evaluating Retrieval Augmented Generation in Competitive Examinations" 1st International Conference on Applied Artificial Intelligence and Machine learning, AIP Conference Proceedings 1. 1(1).

Volumetric Images of Patient Specific Implants for Cranial Defects

Oswal, Dhiren; Khanter, Yugal; Mahalle, Parikshit (2024), " Volumetric Images of Patient Specific Implants for Cranial Defects" Mendeley Data, V1, doi: 10.17632/ywnjr7gnty.1.. 2(2).

Teaching

• Teacher’s Assistant: Coherenece Academy
  • Tutored High school students at a prep school for Indian entrance examination over High school level Physics, Chemistry and Mathematics. Also, graded test paper, took special sessions to improve class average score.

Service and leadership

• Club President, ROTARACT CLUB OF PUNE CAMP NEXT GEN (IND)
  • Community service, Fundraising, Business Facilitations. Jist of major projects:
  • Hold My Hands: Arranged motivational field visit to the National Defence Academy, India, for young orphans by raising INR 60,000 ~$780
  • Mahadaan 8.0: Conducted blood donation camp with a platelet awareness campaign in collaboration with four clubs by collecting 352 blood units and helped hospitals gain at least 10 new platelet donors
  • Passion Junction/Abhipray Incorporated professional networks by conducting mass coffee meetup fair to facilitate a business of INR 350,000 ~$40,000 within club members.