Jun 2022 – Present · 4 yrs · Milan, Lombardy, Italy · On-site

• All research was carried out at Human Technopole in Milan, Italy. During my PhD, I developed four deep-learning–based computational multiplexing (image decomposition) techniques, some of which were published at #ECCV24 and #ICCV23. These methods were designed to improve the imaging of multiple cellular structures within a single fluorescent channel, enabling faster acquisition and reducing photon exposure. I also discovered issues with SSIM on microscopy data and developed an alternative SSIM variant (BIC workshop, ECCV24). Additionally, I contributed in integrating my PhD work to CAREamics (a Pytorch library), created predtiler (a pip installable PyTorch-based tiled prediction tool), and made few contributions to microscopy softwares (c-mda-engine, a C++ microscope control system with SWIG bindings and in microsim, a microscopy simulator).

Feb 2020 – Apr 2021 · 1 yr 2 mos · Taipei City, Taipei City, Taiwan · On-site

• I conducted research on two computer vision problems — (a) gaze estimation in the wild and (b) precipitation nowcasting for Taiwan — both resulting in publications.

Jan 2019 – Feb 2020 · 1 yr 1 mo · Taipei City, Taipei City, Taiwan · On-site

• Self motivated exploration of deep learning field. Did 5 Coursera certifiable courses targeted on specific deep learning domains including but not limited to computer vision, reinforcement learning and bayesian methods. Participated in 4 Kaggle competitions. Was in the top 2-3 percent in the last 2.

Dec 2015 – Dec 2018 · 3 yrs · Greater Bengaluru Area

• Optimized price modeling for portfolio: Development of autoencoder based market neutral strategy. Generated synthetic data to aid in training. It managed 5% of the portfolio.(Python)
• Optimized execution: Development and analysis of multiple intraday execution algorithms and meta algorithms. Used regularized LR and traditional trading techniques like mean reversion, momentum. Daily, $50K was traded using my algorithms saving 1-2 bps. (Python,C++)
• Non ML projects involved
• Data pipeline: Extraction and distributed processing of data from tick data files and web apis. Used in developing models.
• Creation of Execution pipeline and Reconciliation pipeline for multiple brokers : APEX, CQG, IB and TD (Python,C++).
• Conversion of sequential simulation engine to vectorized simulation engine. Achieved 5x speedup (Python).
• Creation of Order Routing Server (ORS) which was the backbone of trading pipeline (C++ & Python).

May 2015 – Oct 2015 · 5 mos · Greater Delhi Area

• Linking news with communities: Scraping news content and using NLP to generate tags from it. Post processing tags to get related communities (Ruby) .
• Recommendation module for rooms and books based on user's facebook like content (Ruby).

May 2013 – Jul 2013 · 2 mos · Development of a proto Emulator in a Discrete Event Network Simulator

• Capture,Creation and Parsing of Raw Packets. Packet Scheduling from real traffic into simulator
• and vice versa and Time Synchronization of real clock with simulator clock.
• Added a minimal Emulator Module in NetSim and successfully tested on internal network.