Working within a small, inter-disciplinary team to bring a RISC-V + Accelerator test chip to life⇢ The test chip serves as a proof-of-concept of a scalable analytics IP architecture⇢ It aims to be "future-proof", capable of running DNN models of today and the future (clairvoyance required)I am among the software folks on the team, tasked with the following migrating the existing software stack from a custom quantization scheme and graph compiler to one powered by open source… Show more Working within a small, inter-disciplinary team to bring a RISC-V + Accelerator test chip to life⇢ The test chip serves as a proof-of-concept of a scalable analytics IP architecture⇢ It aims to be "future-proof", capable of running DNN models of today and the future (clairvoyance required)I am among the software folks on the team, tasked with the following migrating the existing software stack from a custom quantization scheme and graph compiler to one powered by open source tools (PyTorch quantization and Apache TVM).My personal contributions to date:⇢ Initial TVM bringup, implementing full compilation flows from PyTorch models to C code to executable⇢ Development of TVM compilation utility, specialized for the C code generative use-case⇢ “Bring Your Own Codegen” (BYOC) implementation to bring TVM support to our custom accelerator and kernels ⇢ Development of an automated compiler testing framework which can perform the following at scale: ⇢ Inference, in PyTorch, on arbitrary collections of pre-trained models ⇢ Compilation of the PyTorch model to C code to executable ⇢ Inference, of the executable, against the same data inferred upon in PyTorch ⇢ Statistical similarity comparison between the PyTorch and executable output tensors Show less

Lead developer of the TI mmWave Sensing Estimator, working with a team of 3Raw data capture via MPSSE FTDI via SPIE2E Forum customer support Implemented a vector embedding database of relevant E2E forum posts and radar documentation paired with OpenAI APIs for AI-powered responses

Developed a large version update of the mmWave Sensing Estimator, a customer web GUI tool used over 90,000 times, to easily generate and modify configuration parameters for Industrial mmWave Radar DevicesAdvanced a package to export raw radar data over SPI to compensate for a shortage of DCA1000 capture cardsRefined technical communication skills by answering Engineer-to-Engineer (E2E) Forum posts

Applications Engineering Intern under the General Purpose Operational Amplifiers Unit at TI. Built a suite of automation tools to streamline model patching, testing, and validation for PSpice for TI, saving a year's worth of manual model patching. The final evolution of these tools was a full GUI for custom benchmark-testing, design, and automation that works seamlessly with the PSpice for TI Program by sending test netlists to the simulation engine.

Worked under Dr. Swarup Bhunia on (Project 1) developing Nuclear Quadruple Resonance (NQR) techniques to validate the chemical composition of foods, supplements, and medicines, using statistical analysis and machine learning in MATLAB and (Project 2) developing a 3D printing technique where an invisible “tag” or signature can be inserted into an object and detected using an NQR spectrometer as an intrinsic barcode.