Sandeep Dasgupta
Senior Software Engineer at Google
California, United States
Summary
🤩
Rockstar🎓
Top SchoolSandeep Dasgupta is a Senior Software Engineer with 12 years of experience specializing in compilers, MLIR/LLVM ecosystems, and backend compiler work for ML frameworks. Based at Google in California, he has driven integration and verification efforts across StableHLO, MHLO and XLA, contributing notable fixes to collective ops and tuple handling that improve the ML compilation pipeline. His background spans deep program analysis—static, symbolic and dynamic—alongside language design and auto-parallelizing compiler work for both high-level dynamic languages and x86 binaries. Comfortable in Go, C++, Swift, Haskell and low-level C/assembly, he blends research rigor (PhD from UIUC) with production-grade engineering across TensorFlow, JAX and OpenXLA. An active open-source contributor, he implemented and tested many StableHLO ops in reference interpreters and helped modernize MHLO lowering in JAX, work that materially affects how Python+NumPy programs are lowered to accelerators. Colleagues describe him as someone who pairs deep formal thinking about language semantics and verification with pragmatic fixes that keep large ML compiler stacks robust and forward-compatible.
12 years of coding experience
11 years of employment as a software developer
B.E CST, B.E CST at IIEST, ShibpurIndian Institute of Technology KanpurKendriya Vidyalaya Sangathan
English
Stackoverflow
39reputation
33kreached
2answers
3questions
Github Skills (32)
c-language10
stable10
back-end-development10
operation10
python10
datatypes10
tensorrt10
machine-learning10
interpreter10
mlr10
tensorflow10
compiler-optimization10
xla10
compiler10
data-handling10
Programming languages (10)
Github contributions (5)
openxla/stablehlo
Aug 2022 - Jan 2023
Backward compatible ML compute opset inspired by HLO/MHLO
Role in this project:
Back-end Developer
Back-end DeveloperContributions:1970 reviews, 75 commits, 473 PRs in 5 months
Contributions summary:Sandeep implemented and tested several StableHLO operations within the reference interpreter. These operations included `add`, `ceil`, `floor`, `reshape`, `transpose`, `maximum`, `minimum`, `or`, `and`, and various bitwise operations. They also contributed to the handling of different data types and ensuring index compatibility within the tensor implementation. The user's work primarily involved extending the interpreter's functionality to support a broader set of StableHLO ops.
hlomachine-learningcompute
openxla/xla
Dec 2021 - Jan 2023
A machine learning compiler for GPUs, CPUs, and ML accelerators
Role in this project:Back-end Developer
Back-end DeveloperContributions:7 reviews, 23 commits, 16 comments in 1 year 1 month
Contributions summary:Sandeep primarily contributed to the XLA compiler, specifically focusing on the MHLO (MLIR-HLO) dialect and its integration with other components. Their work involved removing tuples from MHLO operations like `whileOp`, `If/Case`, `Infeed/Outfeed/Send/Recv`, and improving the verification of collective communication operations such as `AllGather`, `AllReduce`, and `AllToAll`, indicating a focus on compiler optimizations and improvements to the intermediate representation. They also worked on integrating StableHLO, replacing StrEnumAttr with EnumAttr, and resolving shape-related issues.
compilercommunity-drivenmachine-learningmodular
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.