Abhishek
Enaguthi
I live in the layer underneath the framework: compiler backends, kernel-level debugging, distributed systems at HPC scale.
M.S. in Artificial Intelligence (Thesis) & B.S. Computer Science (AI), Minor in Finance — Oregon State University, expected June 2026.
AI Research Engineer
Research Advanced Computing, Oregon State University
Secured a $1M joint research grant from AMD and IBM for heterogeneous GPU compiler tech. Published compiler research enabling ROCm cross-compilation to x86. Led a team to SC25, placing top 3 of 50+ teams in the IndySCC student cluster competition.
LLM Research Intern
University of Washington
Built multi-turn synthetic QA generation for NVIDIA NeMo and deployed LLaMA-3 70B and Mistral inference via vLLM on NVIDIA H200 clusters — production RAG pipelines on a 50-node, 20TB system for 20+ staff.
Systems Software Engineer
College of Earth Science, Oregon State University
Presented a physics-based digital twin of data centers at Supercomputing '24 using NVIDIA Omniverse and Isaac Sim. Wrote LLVM passes, kernel optimizations, and custom backends — a 30% boost to ML and data processing performance.
HPC Performance Engineer
Oregon State University
Automated SLURM RPC and disk-usage monitoring, adopted by 100+ users daily across OSU's HPC. Benchmarked NCCL and MPI to land 20% faster collective communication and 15% higher memory throughput.
Mocki — Multi-Agent Interview Coaching
Coordinator → Interviewer → Evaluator → Report Generator, adaptive difficulty via the NVIDIA NIM API. 1st place, NVIDIA Nemotron track at BeaverHacks 2026 (RTX 5070 prize), out of 50+ teams.
bluestart — ROCm on POWER9
Configuring ROCm for compatibility with the IBM POWER9 architecture, enabling CUDA code to run on AMD GPUs via HIPify.
compile-your-language
A from-scratch programming language implementation, with an educational guide walking through modern compiler techniques.
TreeHacks — GenAI Research Assistant
A GenAI-powered assistant for academic research, using InterSystems IRIS vector search to surface and cite relevant papers.
Retrieval-Augmented Generation Pipeline
Scaled embedding generation to 235K embeddings from 65K Wikipedia articles using PyTorch DDP across multi-GPU nodes.
{{ activeGalleryPhoto.caption }}
If you are building compilers, GPU systems, or where HPC meets AI, I am listening.
{{ activeBlogPost.title }}
{{ paragraph }}
{{ spotifyTitle }}
{{ spotifyDescription }}
{{ todayTitle }}
{{ todaySummary }}
Welcome to the desktop.
Open apps from the icons or dock. Drag titlebars, resize corners, and use Terminal commands like projects, blog, and gallery.