I'm an ML Systems Research Intern at Embedded LLM, Singapore. I write GPU kernels that make biological and medical foundation models run orders of magnitude faster, on NVIDIA and AMD alike. This year that turned into 8 papers, 3 posters, and 2 accepted talks across 11 venues. Next up: a spotlight talk at NeurIPS in Sydney.
my poster at COLM '26
the badge, say hi
NeurIPS2026 · ASCI Workshop
A deployment card for cancer pathology foundation models, in the units a hospital budgets in.
NeurIPS2026 · TAE Workshop
Auditing how we decide which clinical foundation model to deploy in cancer pathology.
MOML @ MIT2026 · Molecular ML Conference
Profiling-driven kernel fusion for TensorNet in molecular simulation.
ICML2026 · SD4H Workshop
Fused Triton kernels for state-space models in real-time ICU monitoring.
ISCA2026 · HotInfra Workshop · Talk
An infrastructure view of biological foundation model inference across NVIDIA and AMD.
COLM2026 · DAIH Workshop
An energy-aware, cross-vendor protocol for on-premises healthcare AI.
ICPP2026 · DC4AI Workshop
Fused attention compression for long-context genomic foundation models across heterogeneous GPUs.
EurIPS2025 · SimBioChem Workshop
Accelerating molecular simulations with Triton: fused kernels for TensorNet neural potentials.
ACM ICS2026 · Arch4Health · TalkFrom 805 ms to 23 ms: accelerating state-space models for bedside monitoring.
MLSys2026 · Young Professionals Symposium
Portable cross-vendor GPU kernels for high-throughput bioinformatics via OpenAI Triton.
RECOMB2026 · ARCH
Fused kernels for high-throughput biological foundation models, on any vendor.
ISC HPC2026 · Hamburg
Biological foundation models and algorithms accelerated with OpenAI Triton.
Selected for a spotlight talk.
364 Slides a Shift, Not 222 accepted.
Student delegate at Google Ananta, Oct 29–30.
Change the Metric, Change the Winner accepted.
From 13 Hours to 4.6 accepted.
Fused attention compression for genomic models accepted.
On-premises clinical AI protocol accepted.
Invited to review for both workshops.
When the LLM-Tuned Stack Misses accepted as a short talk.
From 805 ms to 23 ms accepted.
BioTriton accepted.
ICU monitoring work accepted as a short talk.
Hardware-portable fused kernels accepted.
Portable Triton kernels accepted.
TensorNet kernels accepted, and invited to review.
Joined as an ML Systems Research Intern.
ML Systems Research Intern · Singapore · Remote
Software Engineering Intern · North Carolina, USA
Freelance Machine Learning Engineer · India · Freelance
B.E. Computer Engineering · Patiala, India