May 2025 – Feb 2026 · 9 mos · Los Angeles, California, United States · Hybrid

• Topic: 𝗔 𝗗𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁𝗶𝗮𝗹𝗹𝘆 𝗣𝗿𝗶𝘃𝗮𝘁𝗲 𝗢𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗔𝗽𝗽𝗿𝗼𝗮𝗰𝗵 𝗳𝗼𝗿 𝗟𝗮𝘆𝗲𝗿 𝗪𝗶𝘀𝗲 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗶𝗻 𝗟𝗟𝗠𝘀
• 𝗖𝘂𝘁 𝗟𝗟𝗠 𝘁𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗼𝘃𝗲𝗿𝗵𝗲𝗮𝗱 𝗯𝘆 𝟭𝟴% using zeroth-order optimization methods for scalable, privacy-preserving deployment.
• 𝗜𝗻𝗰𝗿𝗲𝗮𝘀𝗲𝗱 𝗺𝗲𝗺𝗼𝗿𝘆 𝗲𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝗰𝘆 𝗯𝘆 𝟭𝟮% with lightweight optimizers, enabling industry-scale model training.
• 𝗕𝘂𝗶𝗹𝘁 𝗽𝗿𝗶𝘃𝗮𝗰𝘆-𝗳𝗶𝗿𝘀𝘁 𝗼𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗽𝗶𝗽𝗲𝗹𝗶𝗻𝗲𝘀 to reduce compute cost without sacrificing model utility.

May 2025 – Jan 2026 · 8 mos · Los Angeles, California, United States · On-site

• Selected among 𝟭𝟱 𝘀𝘁𝘂𝗱𝗲𝗻𝘁𝘀 𝗮𝗰𝗿𝗼𝘀𝘀 𝗜𝗻𝗱𝗶𝗮 for the prestigious 𝗜𝗻𝗱𝗼-𝗨𝗦 𝗦𝗰𝗶𝗲𝗻𝗰𝗲 & 𝗧𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝘆 𝗙𝗼𝗿𝘂𝗺 𝗙𝗲𝗹𝗹𝗼𝘄𝘀𝗵𝗶𝗽.
• In collaboration with 𝗠𝗲𝘁𝗮, worked on 𝗲𝗻𝗵𝗮𝗻𝗰𝗶𝗻𝗴 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁𝗶𝗮𝗹 𝗽𝗿𝗶𝘃𝗮𝗰𝘆 𝗶𝗻 𝗹𝗮𝗿𝗴𝗲 𝗹𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗺𝗼𝗱𝗲𝗹𝘀 𝘁𝗵𝗿𝗼𝘂𝗴𝗵 𝗿𝗲𝘀𝗼𝘂𝗿𝗰𝗲-𝗲𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝘁 𝗼𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝘁𝗲𝗰𝗵𝗻𝗶𝗾𝘂𝗲𝘀. The project focused on advancing the 𝘁𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗮𝗻𝗱 𝗺𝗲𝗺𝗼𝗿𝘆 𝗲𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝗰𝘆 𝗼𝗳 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁𝗶𝗮𝗹𝗹𝘆 𝗽𝗿𝗶𝘃𝗮𝘁𝗲 𝗼𝗽𝘁𝗶𝗺𝗶𝘇𝗲𝗿𝘀 used in large language models (LLMs), aiming to make privacy-preserving AI more scalable and practical for real-world deployment.

Apr 2025 – Jan 2026 · 9 mos · Stanford, California, United States · Hybrid

• Topic: 𝗔𝗜 𝗔𝗱𝗼𝗽𝘁𝗶𝗼𝗻 𝗶𝗻 𝗘𝗺𝗲𝗿𝗴𝗶𝗻𝗴 𝗠𝗮𝗿𝗸𝗲𝘁𝘀: 𝗕𝗮𝗹𝗮𝗻𝗰𝗶𝗻𝗴 𝗕𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝗜𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻 𝘄𝗶𝘁𝗵 𝗛𝘂𝗺𝗮𝗻-𝗖𝗲𝗻𝘁𝗿𝗶𝗰 𝗢𝘂𝘁𝗰𝗼𝗺𝗲𝘀.
• Developed an ML decision-making framework 𝗶𝗺𝗽𝗿𝗼𝘃𝗶𝗻𝗴 𝗽𝗿𝗲𝗱𝗶𝗰𝘁𝗶𝘃𝗲 𝗮𝗰𝗰𝘂𝗿𝗮𝗰𝘆 of workforce adoption models 𝗯𝘆 𝟳%.
• Quantified productivity trade-offs of AI integration, identifying 𝟮𝟵% variance across industry sectors.
• 𝗗𝗲𝘀𝗶𝗴𝗻𝗶𝗻𝗴 𝗮 𝗱𝗲𝗰𝗶𝘀𝗶𝗼𝗻-𝘀𝘂𝗽𝗽𝗼𝗿𝘁 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 combining ML outputs with ROI metrics to evaluate adoption scenarios.
• Collaborated with economists to translate ML insights into policy-oriented recommendations for responsible AI use.

Nov 2024 – Apr 2025 · 5 mos · Ithaca, New York, United States · Remote

• Topic: 𝗜𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗻𝗴 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗠𝗼𝗱𝗲𝗹𝘀 𝘄𝗶𝘁𝗵 𝗕𝗶𝗼𝗺𝗲𝗱𝗶𝗰𝗮𝗹 𝗞𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 𝗚𝗿𝗮𝗽𝗵𝘀 𝗳𝗼𝗿 𝗜𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝘁 𝗗𝗮𝘁𝗮 𝗤𝘂𝗲𝗿𝘆𝗶𝗻𝗴
• 𝗗𝗲𝗽𝗹𝗼𝘆𝗲𝗱 𝗮𝗻 𝗡𝗟𝗣-𝗱𝗿𝗶𝘃𝗲𝗻 𝗾𝘂𝗲𝗿𝘆 𝗶𝗻𝘁𝗲𝗿𝗳𝗮𝗰𝗲, reducing clinician 𝗶𝗻𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻 𝗿𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹 𝘁𝗶𝗺𝗲 𝗯𝘆 𝟰𝟯%.
• Achieved 𝟴𝟲% 𝗾𝘂𝗲𝗿𝘆 𝘀𝘂𝗰𝗰𝗲𝘀𝘀 𝗿𝗮𝘁𝗲 via 𝗴𝗿𝗮𝗽𝗵-𝗯𝗮𝘀𝗲𝗱 𝗺𝘂𝗹𝘁𝗶-𝗵𝗼𝗽 𝗿𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴 for drug–disease interaction queries.
• Engineered a scalable 𝗟𝗟𝗠–𝗸𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 𝗴𝗿𝗮𝗽𝗵 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 for drug discovery and decision support.

Sep 2024 – Apr 2025 · 7 mos · Princeton, New Jersey, United States · Remote

• Topic: 𝗠𝗲𝗺𝗼𝗿𝘆 𝗨𝘀𝗮𝗴𝗲 𝗢𝗽𝘁𝗶𝗺𝗶𝘇𝗶𝗻𝗴 𝗳𝗼𝗿 𝗡𝗲𝘂𝗿𝗮𝗹 𝗡𝗲𝘁𝘄𝗼𝗿𝗸 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗼𝗻 𝗟𝗼𝘄-𝗥𝗲𝘀𝗼𝘂𝗿𝗰𝗲 𝗗𝗲𝘃𝗶𝗰𝗲𝘀
• 𝗥𝗲𝗱𝘂𝗰𝗲𝗱 𝗺𝗲𝗺𝗼𝗿𝘆 𝘂𝘀𝗮𝗴𝗲 𝗯𝘆 𝟮𝟭% and 𝘁𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝘁𝗶𝗺𝗲 𝗯𝘆 𝟭𝟲% on Jetson Nano and Raspberry Pi for DL models.
• Achieved 𝟵𝟯% (CIFAR-10) and 𝟵𝟴% (MNIST) accuracy by fine-tuning 𝗠𝗼𝗯𝗶𝗹𝗲𝗡𝗲𝘁𝗩𝟭 and 𝗥𝗲𝘀𝗡𝗲𝘁-𝟭𝟴.
• Built a 𝗺𝗼𝗱𝘂𝗹𝗮𝗿 𝘁𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗽𝗶𝗽𝗲𝗹𝗶𝗻𝗲 𝗳𝗼𝗿 𝘀𝗰𝗮𝗹𝗮𝗯𝗹𝗲 𝗱𝗲𝗽𝗹𝗼𝘆𝗺𝗲𝗻𝘁 of memory-efficient ML models on edge AI devices.

Jul 2024 – Jul 2024 · 0 mo · Remote

May 2024 – Sep 2024 · 4 mos

Feb 2024 – Aug 2024 · 6 mos · New York, United States · Remote

• Topic: 𝗙𝗶𝗻𝗲-𝗧𝘂𝗻𝗶𝗻𝗴 𝗟𝗮𝗿𝗴𝗲 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗠𝗼𝗱𝗲𝗹𝘀 𝘄𝗶𝘁𝗵 𝗥𝗲𝗶𝗻𝗳𝗼𝗿𝗰𝗲𝗺𝗲𝗻𝘁 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗳𝗼𝗿 𝗗𝗶𝗮𝗹𝗼𝗴𝘂𝗲 𝗦𝘆𝘀𝘁𝗲𝗺𝘀
• Reduced incoherent responses by 𝟭𝟲% with a scalable 𝗿𝗲𝗶𝗻𝗳𝗼𝗿𝗰𝗲𝗺𝗲𝗻𝘁 𝗹𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗽𝗶𝗽𝗲𝗹𝗶𝗻𝗲 for dialogue systems.
• Improved dialogue coherence by 𝟭𝟭% via 𝗳𝗶𝗻𝗲-𝘁𝘂𝗻𝗶𝗻𝗴 𝗚𝗣𝗧-𝟯.𝟱 on DailyDialog and custom datasets.
• Built a reusable 𝗲𝘃𝗮𝗹𝘂𝗮𝘁𝗶𝗼𝗻 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 (BLEU, perplexity, diversity) to standardize benchmarking.

Sep 2023 – Dec 2023 · 3 mos

Aug 2023 – Aug 2024 · 1 yr · Bengaluru, Karnataka, India · Remote

• Topic: 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 𝗮𝗻𝗱 𝗔𝗻𝗮𝗹𝘆𝘀𝗶𝘀 𝗼𝗳 𝗟𝗶𝗴𝗵𝘁𝘄𝗲𝗶𝗴𝗵𝘁 𝗢𝗽𝘁𝗶𝗺𝗶𝘇𝗲𝗿𝘀 𝗳𝗼𝗿 𝗦𝗰𝗮𝗹𝗮𝗯𝗹𝗲 𝗡𝗲𝘂𝗿𝗮𝗹 𝗡𝗲𝘁𝘄𝗼𝗿𝗸 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴
• Reduced 𝗚𝗣𝗨 𝗵𝗼𝘂𝗿𝘀 𝗯𝘆 𝟮𝟮% using lightweight optimizers while preserving accuracy.
• Accelerated training by 𝟭𝟳% on MNIST and CIFAR-10 with grid search and Bayesian optimization.
• Improved 𝗺𝗼𝗱𝗲𝗹 𝘀𝘁𝗮𝗯𝗶𝗹𝗶𝘁𝘆 𝗯𝘆 𝟭𝟯%, highlighting trade-offs between accuracy and efficiency in deployment.

Jun 2023 – Aug 2023 · 2 mos · Mumbai, Maharashtra, India · On-site

• Topic: 𝗟𝗲𝘃𝗼𝗱𝗼𝗽𝗮-𝗜𝗻𝗱𝘂𝗰𝗲𝗱 𝗗𝘆𝘀𝗸𝗶𝗻𝗲𝘀𝗶𝗮 𝗣𝗿𝗲𝗱𝗶𝗰𝘁𝗶𝗼𝗻 𝗳𝗿𝗼𝗺 𝗦𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗮𝗹 𝗠𝗥𝗜 𝗶𝗻 𝗣𝗮𝗿𝗸𝗶𝗻𝘀𝗼𝗻’𝘀 𝗗𝗶𝘀𝗲𝗮𝘀𝗲 𝘂𝘀𝗶𝗻𝗴 𝗠𝗮𝗰𝗵𝗶𝗻𝗲 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴
• 𝗔𝗰𝗵𝗶𝗲𝘃𝗲𝗱 𝟵𝟲% 𝗮𝗰𝗰𝘂𝗿𝗮𝗰𝘆 𝗮𝗻𝗱 𝟬.𝟵𝟴 𝗔𝗨𝗖 by developing a CNN-based deep learning pipeline to predict Levodopa-induced dyskinesia from MRI scans.
• 𝗘𝗻𝗵𝗮𝗻𝗰𝗲𝗱 𝗱𝗶𝗮𝗴𝗻𝗼𝘀𝘁𝗶𝗰 𝘀𝗶𝗴𝗻𝗮𝗹 𝗲𝘅𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻 through voxel-based morphometry and segmentation, identifying clinically relevant brain regions.
• 𝗕𝘂𝗶𝗹𝘁 𝗮 𝗿𝗲𝗽𝗿𝗼𝗱𝘂𝗰𝗶𝗯𝗹𝗲 𝗺𝗲𝗱𝗶𝗰𝗮𝗹 𝗶𝗺𝗮𝗴𝗶𝗻𝗴 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 supporting data-driven decision-making in Parkinson’s disease management.

Mar 2023 – Jul 2023 · 4 mos · Roorkee, Uttarakhand, India · Remote

• Topic: 𝗠𝗮𝗰𝗵𝗶𝗻𝗲 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴-𝗗𝗿𝗶𝘃𝗲𝗻 𝗣𝗿𝗲𝗱𝗶𝗰𝘁𝗶𝘃𝗲 𝗠𝗼𝗱𝗲𝗹𝗶𝗻𝗴 𝗳𝗼𝗿 𝗘𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝘁 𝗠𝗮𝘁𝗲𝗿𝗶𝗮𝗹 𝗗𝗲𝘀𝗶𝗴𝗻 𝗼𝗳 𝗠𝗫𝗲𝗻𝗲-𝗕𝗮𝘀𝗲𝗱 𝗦𝗸𝗶𝗻-𝗟𝗶𝗸𝗲 𝗘𝗹𝗲𝗰𝘁𝗿𝗼𝗱𝗲𝘀
• 𝗖𝘂𝘁 𝘀𝗶𝗺𝘂𝗹𝗮𝘁𝗶𝗼𝗻 𝘁𝗶𝗺𝗲 𝗳𝗿𝗼𝗺 𝟱𝟬𝟬 𝗖𝗣𝗨 𝗵𝗼𝘂𝗿𝘀 𝘁𝗼 𝟮𝟬 𝗚𝗣𝗨 minutes by applying Gaussian and Symbolic Regression for MXene band edge and stability prediction.
• 𝗔𝗰𝗵𝗶𝗲𝘃𝗲𝗱 𝟵𝟲–𝟵𝟵% 𝗮𝗰𝗰𝘂𝗿𝗮𝗰𝘆 in cytotoxicity prediction, haptic sensor analysis, and pesticide detection using Random Forest + PCA, CNNs, and LSTMs.
• 𝗕𝘂𝗶𝗹𝘁 𝗮 𝘀𝗰𝗮𝗹𝗮𝗯𝗹𝗲 𝗺𝗮𝘁𝗲𝗿𝗶𝗮𝗹𝘀 𝗶𝗻𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗰𝘀 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 integrating simulations and experiments to accelerate MXene-based wearable biosensor design.

Feb 2023 – May 2023 · 3 mos · Hyderabad, Telangana, India · Remote

• Topic: 𝗖𝗼𝗴𝗻𝗶𝘁𝗶𝘃𝗲 𝗡𝗲𝘂𝗿𝗼𝘀𝗰𝗶𝗲𝗻𝗰𝗲 𝗦𝘁𝘂𝗱𝘆 𝗼𝗻 𝗡𝗲𝘂𝗿𝗮𝗹 𝗗𝗿𝗶𝗳𝘁 𝗨𝘀𝗶𝗻𝗴 𝗠𝗮𝗰𝗵𝗶𝗻𝗲 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴
• 𝗥𝗲𝗱𝘂𝗰𝗲𝗱 𝗘𝗘𝗚 𝗻𝗼𝗶𝘀𝗲/𝗮𝗿𝘁𝗶𝗳𝗮𝗰𝘁𝘀 𝗯𝘆 𝟮𝟯% using advanced preprocessing, enabling reliable downstream ML analysis.
• 𝗔𝗰𝗵𝗶𝗲𝘃𝗲𝗱 𝟵𝟳% 𝗮𝗰𝗰𝘂𝗿𝗮𝗰𝘆 in classifying brain states via time–frequency analysis and ML models on EEG patterns.
• 𝗨𝗻𝗰𝗼𝘃𝗲𝗿𝗲𝗱 𝗻𝗲𝘂𝗿𝗮𝗹 𝗱𝗿𝗶𝗳𝘁 𝗽𝗮𝘁𝘁𝗲𝗿𝗻𝘀 with applications in brain–computer interfaces and cognitive monitoring.

Oct 2022 – Feb 2023 · 4 mos · Aligarh, Uttar Pradesh, India · On-site

• Topic: 𝗔 𝗖𝗥𝗡𝗡 𝗔𝗽𝗽𝗿𝗼𝗮𝗰𝗵 𝘁𝗼 𝗥𝗲𝘀𝘁𝗶𝗻𝗴-𝗦𝘁𝗮𝘁𝗲 𝗘𝗘𝗚 𝗖𝗹𝗮𝘀𝘀𝗶𝗳𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝗶𝗻 𝗣𝗮𝗿𝗸𝗶𝗻𝘀𝗼𝗻’𝘀 𝗗𝗶𝘀𝗲𝗮𝘀𝗲
• 𝗕𝘂𝗶𝗹𝘁 𝗮 𝗖𝗥𝗡𝗡 𝗺𝗼𝗱𝗲𝗹 (𝟭𝗗 𝗖𝗡𝗡 + 𝗚𝗥𝗨) to capture spatiotemporal EEG features for Parkinson’s disease detection.
• 𝗢𝘂𝘁𝗽𝗲𝗿𝗳𝗼𝗿𝗺𝗲𝗱 𝗽𝗿𝗶𝗼𝗿 𝗯𝗮𝘀𝗲𝗹𝗶𝗻𝗲𝘀 with 𝟵𝟳.𝟲% 𝗮𝗰𝗰𝘂𝗿𝗮𝗰𝘆, 𝟵𝟲.𝟯% 𝗽𝗿𝗲𝗰𝗶𝘀𝗶𝗼𝗻, 𝗮𝗻𝗱 𝟵𝟴.𝟭% 𝗿𝗲𝗰𝗮𝗹𝗹 on resting-state EEG classification.
• 𝗩𝗮𝗹𝗶𝗱𝗮𝘁𝗲𝗱 𝗘𝗘𝗚-𝗯𝗮𝘀𝗲𝗱 𝗯𝗶𝗼𝗺𝗮𝗿𝗸𝗲𝗿𝘀 for early detection and clinical monitoring of Parkinson’s disease.

Oct 2022 – Dec 2022 · 2 mos