Darya Mikhailenko is a graduate student in Electrical and Computer Engineering at the University of Rochester. She is doing research in computer architecture under the supervision of Prof. Engin Ipek.
Darya is expected to graduate in May 2020 with a Master’s Degree.
Darya holds a Bachelor of Science diploma in Electrical and Electronics Engineering from Nazarbayev University, Kazakhstan. There, she was a part of the Bioinspired Microelectronics Systems Lab research group and worked on modeling neuromorphic architecture for complex cognitive tasks.
Her interests include energy-efficient computer architectures; high-performance, low-power digital VLSI circuit design; SoC performance and power estimation; analysis of power delivery networks; CPU/GPU architecture; machine learning; neuromorphic accelerators; and architecture design for emerging technologies. Her recent paper “Commutative Data Reordering: A New Technique to Reduce Data Movement Energy on Sparse Inference Workloads” was accepted to ISCA 2020, which is the premier forum in computer architecture.
Currently, she is working on developing an algorithm to map the weight matrices of sparse neural networks (NN) onto memristive crossbars for efficient in-situ processing. Executing sparse NNs on ReRAM-based in-situ accelerators results in wasted energy to charge an entire ReRAM crossbar where most elements are zero. Darya proposes to reorder the weights in a way that maximizes the number of all-zero crossbars, which can then be put into a low-power sleep mode. This method will increase the energy efficiency and performance of neuromorphic accelerators without loss of accuracy.
Darya did internships in academia (Purdue University, 2017) and in industry (Google, 2019).
At Purdue, she worked under the supervision of Prof. Kaushik Roy and analyzed the behavior of sneak-path currents in memristive crossbar architectures. Her area- and energy-efficient solution to minimize sneak-path currents “M^2CA: Modular Memristive Crossbar Arrays” was published at ISCAS 2018.
At Google, Darya was a part of the Google Silicon Power Architecture team. She collaborated with hardware designers and architects to build new pre-silicon tools and enhance current frameworks for power and performance estimation of use cases in the mobile space.
Darya is an active-lifestyle supporter. She was a captain of a women’s volleyball team and a lead Zumba dancer. In her free time, she enjoys hiking, cycling, rowing, and swimming.
