- Nov 2025 – Present
ML Researcher @Arizona State University
Building a modular PyTorch Lightning + JAX framework so Applied Materials can stress-test whether neural networks can replace traditional plasma solvers in their etch chambers.
Stack
- PyTorch Lightning
- JAX
- Hydra
- WandB
- Jul 2025 – Oct 2025
ML Engineer Intern @OpenEye, Cadence Design Systems
ESM2 + contrastive-learning pipeline batched through cuDF and PyTorch Lightning. Replaced GP bottleneck with cuML/RAFT — batch sizes jumped 150 → 50K+ at 7.2 ms / sequence.
Stack
- ESM2
- cuDF
- cuML
- RAFT
- OmegaConf
- Jun 2024 – Sep 2024
GenAI Engineering Intern @Talin Labs
Deployed a fine-tuned Mistral-7B-Q8 on Kubernetes behind FastAPI with a six-agent LangChain orchestration. Sub-200 ms p95 for 10K users; 88% on 10K human-evaluated queries.
Stack
- Mistral-7B
- Kubernetes
- FastAPI
- LangChain
- LangGraph
Hi! I’m Rushir
Bhavsar.
Physics-informed neural nets at ASU. Previously Cadence.Currently at ASU, researching physics-informed neural networks for plasma simulation. Previously at Cadence on protein property prediction at million-sequence scale.Open to ML systems / infrastructure roles.Looking for roles in ML systems and ML infrastructure — orchestration, GPU pipelines, and evaluation harnesses.
.png&w=2048&q=75)
+ ML systems / infra
+ Tempe, Arizona
- Master's thesis
MACE-PINN
Parallel-subnetwork PINN with random Fourier feature embeddings for coupled PDE systems. 40–60% lower L2 error than single-network baselines on Gray-Scott and Ginzburg-Landau.
- LLM benchmark
MedQuery
10K-query classification benchmark across four difficulty levels and nine categories — comparing local quantised models (MLX, llama.cpp) against frontier APIs honestly.
Stack
- LangGraph
- LangChain
- FAISS
- OpenAI
Links
- Retrieval system
HybridFlow
Hybrid Qdrant + Neo4j + SQLite retrieval across surgical textbooks with sub-12 ms P50 multi-hop reasoning through ten tool functions.
Stack
- Qdrant
- Neo4j
- SQLite
- FastAPI
Links
- Coursework, prod-quality
Yelp recs & sentiment
Spark ALS recommender plus a sentiment classifier behind FastAPI microservices serving 5M+ interactions at sub-100 ms p95, with Redis caching and MLflow tracking.
Stack
- PySpark
- FastAPI
- MLflow
- Docker
- Redis
Links
- 2025
MS, Data Science
Arizona State University · HPC concentration
Coursework
Statistical ML · Distributed Systems · GPU Computing · Deep Learning Theory · Numerical Methods · Optimization
Highlights
- Thesis on physics-informed neural networks for coupled PDE systems
- GPA 3.71
- 2023
BTech, Computer Science & Engineering
Nirma University, Ahmedabad
Coursework
Algorithms · Operating Systems · Computer Networks · Databases · Compilers · Linear Algebra
Looking for
a role.
Email’s the fastest way to reach me. Especially open to ML systems and infrastructure roles.
- Emailbhavsarrushir@gmail.com↗
- LinkedInrushir-bhavsar↗
- GitHubrushirb2001↗
- ResumePDF↗