2 ms·
Location: San Diego, US Remote: Yes (open to hybrid or in-person) Willing to relocate: Yes (preferred Bay Area, NYC, Austin, and Boston) Visa: F-1 OPT, 3 yea
by prasid74 2mo ago
Location: San Diego, US
Remote: Yes (open to hybrid or in-person)
Willing to relocate: Yes (preferred Bay Area, NYC, Austin, and Boston)
Visa: F-1 OPT, 3 years of US work authorization (STEM OPT eligible)
Technologies: Python, PyTorch, Hugging Face, LLMs, RAG, Qdrant, Hybrid Retrieval (BM25 + dense), Cross-Encoder Reranking, PEFT/LoRA, LangGraph, NLP, Computer Vision, PySpark, Vertex AI, AWS, GCP, Docker, Kubernetes, W&B, Airflow, SQL, C++
Resume/CV: https://kubershahi.github.io https://kubershahi.github.io
Email: kshahi@ucsd.edu
ML Engineer / AI Engineer with 2+ years of industry experience across ML systems, NLP, and large-scale data pipelines. Currently a Graduate Student Researcher at UCSD's BIAG Lab, where I built an end-to-end PyTorch/CUDA pipeline for registration-error regression, training a 3D U-Net that hit Pearson r = 0.874 on 147 held-out volumes. Previously built production Vertex AI training pipelines cutting experiment turnaround by 20% (Melio), and distributed PySpark ETL processing 5M+ invoices/year plus graph analytics driving 15% revenue growth (Vayana Network).
MS CS (AI) at UC San Diego, GPA 4.0. Recent projects include a dual-corpus RAG pipeline for earnings-call contradiction detection with hybrid retrieval and reranking (CrossCheck), an LLM agent evaluation study under adversarial/deceptive conditions across 4 architectures, and low-resource English-Maithili NMT with LoRA fine-tuning. Open to MLE, AI Engineer, Applied Scientist and Research Engineer roles.