Ashutosh Parkhi

CTO

Cambridge, England, United Kingdom15 yrs 5 mos experience

Highly Stable

Key Highlights

Led PyTorch compiler team for optimized ML performance.
Contributed to TVM for efficient neural network execution.
Developed TensorFlow tools for advanced AI applications.

Stackforce AI infers this person is a Machine Learning Compiler Engineer with expertise in AI Tooling and Software Development.

Contact

Skills

Core Skills

Machine LearningCompiler OptimizationProject Delivery

Other Skills

CommunicationSmall Team ManagementSoftware DevelopmentTeam ManagementTeam LeadershipTeam BuildingPython (Programming Language)JenkinsTLMTensorFlowSDKsC++CPerlAlgorithms

About

With a robust background in software solutions for hardware device mapping, my role as an Engineering Lead at Arm involves leading one of the PyTorch compiler teams to enhance machine learning model performance. Our contributions, crucial to the open source ecosystem of PyTorch empower developers to achieve optimized machine learning implementations. Previously, as a Staff Compiler Engineer, I was committed to the TVM apache project, integrating neural network libraries and backend compilers, which enabled efficient heterogeneous execution across devices. The collaborative efforts of our team ensure groundbreaking strides in AI tooling and device backend enablement.

Experience

15 yrs 5 mos

Total Experience

3 yrs 10 mos

Average Tenure

Current Experience

Arm

3 roles

Engineering Lead

Promoted

Mar 2024 – Jan 2025 · 10 mos · Cambridge, England, United Kingdom · Hybrid

Team lead responsible for delivering PyTorch/ExecuTorch compilation for Arm's NPU.
Team lead responsible for delivering TFLite to TOSA (IR for Arm IPs) legalisation.
Line manager and hiring manager for the team.

CommunicationProject DeliveryMachine LearningCompiler Optimization

Technical Team Lead

Jun 2023 – Feb 2024 · 8 mos · Cambridge, England, United Kingdom · Hybrid

-Technical lead for designing Keras3 to IREE flow to map mixed ML and CV applications.

Small Team ManagementProject DeliveryMachine LearningCompiler Optimization

Staff Compiler Engineer

May 2021 – May 2023 · 2 yrs · Cambridge, England, United Kingdom · Hybrid

As a TVM apache committer, I have contributed to integration and maintenance of multiple neural network (NN) software libraries and backend compilers. TVM allows mapping of NN workloads onto these IPs while allowing efficient heterogeneous execution across multiple devices.

Machine LearningSoftware Development

Imagination technologies

Senior Software Engineer

Nov 2017 – Apr 2021 · 3 yrs 5 mos · Pune, Maharashtra, India · On-site

Developed Tensorflow equivalents for network partitioning, graph transforms, quantization tools in TVM. Designed a heterogeneous flow in NNVM to support Imagination's GPU and NNA (Neural network accelerator). Delivered a complex SDK involving multiple work flows supporting GPU and NNA.

Machine LearningSoftware Development

Intel corporation

Simulation/Modeling Engineer

Jun 2012 – Nov 2017 · 5 yrs 5 mos · Bengaluru, Karnataka, India

Developed functional simulation model for a media accelerator. Brought up a Face Recognition application on an integrated CPU-GPU simulation system. Developed SystemC/TLM2 models for Intel's neural networks accelerators to enable firmware bring up and IP verification.

Machine LearningSoftware Development

Mentor graphics

Senior Member Technical Staff

Jun 2007 – Jul 2010 · 3 yrs 1 mo · Noida, Uttar Pradesh, India

Developed a TCL/TK UI for editing combination of RTL signals for debugging the chip designs post Veloce emulation. Designed converters for Veloce to debussy/vcd trace formats to standardize signal viewing.

Software Development