Heterogeneous architectures for accelerating ultra-long sequence alignment

Description

This project focuses on designing a custom, domain-specific accelerator to accelerate sequence alignment tasks by offloading the computationally and memory-intensive parts to specialized hardware. The approach involves modeling and simulating hybrid architectures using GEM5, then implementing RTL designs, verifying them, and moving towards physical synthesis and design validation. Its component bridges general-purpose processing and hardware acceleration, leveraging bit-parallelism and dynamic programming optimizations tailored to genomic alignment workloads.

Background

Oscar completed his degrees in Computer Engineering and Telecommunications Electronics Engineering and later pursued a Master in Advanced Telecommunications Technologies, where he specialized in digital design of embedded systems and hardware acceleration. Oscar has extensive experience in hardware architecture, digital design and simulation environments. He has worked with RISC-V architectures and designed accelerators using RTL development flows, hardware modeling and verification. His experience allows him to navigate the full stack, from high-level algorithmic vision to hardware implementation.

Motivation

Today's bioinformatics workloads, particularly the alignment of ultra-long DNA sequences, demand computational frameworks that conventional CPUs or GPUs struggle to handle efficiently. Oscar is motivated by the challenge of pushing the boundaries of custom hardware accelerators, reconciling algorithmic innovation with physical implementation. He seeks to make high-throughput genomic analysis more efficient, reducing cost, power, and latency, thereby facilitating advances in precision medicine and large-scale genomics. Oscar's research interests include heterogeneous architectures, RISC-V architectures, domain-specific hardware design, and hardware accelerators for genomic alignment.