Enabling cost effective AI Inference at scale

Compiler Engineer in Qualcomm's AI Research division working on developing optimising compilers for mapping neural networks to custom hardware.

Postdoctoral researcher at the Institute for Computing Systems Architecture, School of Informatics under the supervision of Dr. Tobias Grosser. Worked on developing an intermediate representation that allows to better map high level C code to dynamically scheduled FPGA circuits.

My current research interests are; Enabling on-device learning to allow Convolutional Neural Networks adapt to the data it is processing in a server-free manner, Low-precision CNN training techniques and FPGA hardware architectures for CNN training. I have experience in using PyTorch for GPU based develoment and Vivado HLS for FPGA based development. 4 peer-reviewed publications:- [ICCV 2021] perf4sight: A toolflow to model CNN training performance on Edge GPUs (First Author)- [ICML 2020] Multi-precision Policy Enforced Training (MuPPET): A precision-switching strategy for quantised fixed-point training of CNNs (First Author)- [CVPR 2020] Now that I can see, I can improve: Enabling data-driven finetuning of CNNs on Edge Devices (First Author)- [FPL 2020] Caffe Barista: Brewing Caffe with FPGAs in the Training Loop (Second Author)

6-month compiler research Intern working on developing an LLVM compiler backend for the CPU architecture research team in order to facilitate architecture prototyping.

Worked for 6 months as an embedded hardware and software engineer in a marine seismic acquisition team. Technology covered include : - PCB design using KiCad - Amplifier design and simulation using LtSpice - PCB vibration testing - MicroBlaze softcore processor by Xilinx - Xilinx Zynq SoC - FPGA design

Working in the Intelligent Networks and Systems research group, researching Self-organising Community Energy Systems. Responsibilities include: - Developing a Raspberry-Pi demonstrator of a smart house by connecting various peripherals to it and integrating them using python code- iOS app development using objective-C which works in tandem with the Raspberry-Pi demonstrator to give users up to date information about their smart house and also allow them to control appliances- Web server deployment using Java to allow the app and the Raspberry pi to communicate - Running a simulation of multi-agent systems using the Java based Presage2 software

- Developed software using java script and python intended to commercially be used for stock trading- Worked with no-sql databases such as cassandra and redis - Developed software to be used for both user interface as well as backend database operations