Research Scientist (Multimodal Generative AI)

John Doe

Multimodal generative AI researcher applying LLM evaluation and multimodal methods to real-world and biological problems.

Actively Searching

John Doe

About

I am a multimodal generative AI researcher and Ph.D. candidate at Columbia University focused on LLMs, multimodal learning, and human–AI interaction. I build and evaluate large-scale transformer models and scalable training/evaluation pipelines to improve model reliability and reduce hallucinations, and I am now applying these methods toward biological AI. I have industry experience at OpenAI and a DeepMind research internship collaborating across product, safety, and infrastructure teams.

Looking Ahead

I apply multimodal generative AI and rigorous LLM evaluation methods to solve real-world problems. I am now focusing on translating these techniques to biological AI applications.

Skills

Languages
Python,
C++,
Java,
SQL,
JavaScript
Frameworks
PyTorch,
TensorFlow,
Hugging Face,
LangChain
Cloud & Tools
AWS,
Docker,
Kubernetes,
Linux,
Git
Research Areas
LLMs,
NLP,
Reinforcement Learning,
Computer Vision,
Multimodal Learning,
Generative AI,
Human-AI Interaction,
Knowledge Retrieval,
Document Understanding
Methods & Concepts
Evaluation & Metrics,
Retrieval-Augmented Generation (RAG),
Long-context Reasoning
Systems & Infrastructure
Distributed Training,
Scalable Evaluation Pipelines
Models
Transformer Models
Safety & Alignment
AI Alignment / Model Safety
Deployment & Production
Production AI Systems
Professional Skills
Mentoring,
Cross-functional Collaboration,
Research Leadership,
Experimental Research & Evaluation Design

Selected Projects

Multimodal Evaluation Framework to Reduce Hallucinations

Reduced hallucination rates by 22% and improved overall model reliability by implementing comprehensive evaluation frameworks that combined safety checks, automated metrics, and human-in-the-loop validation. Frameworks were integrated with production pipelines to enable continuous monitoring and informed model improvements.

Python · Hugging Face · LangChain · Docker · Kubernetes · AWS

Scalable Generative AI Evaluation Pipeline

Enabled processing of millions of samples daily to validate generative models by building scalable evaluation pipelines that supported experiments in retrieval-augmented generation and long-context reasoning. Delivered reproducible infrastructure used to surface findings for senior research leadership.

Python · Docker · Kubernetes · Linux · Git

Distributed Training Pipeline for Large-Scale Transformer Models

Accelerated research throughput and enabled training of large-scale transformer models by designing distributed training pipelines used in multimodal document understanding work that led to publications at NeurIPS and ACL. Pipelines supported efficient resource utilization and collaboration across graduate researchers.

PyTorch · Kubernetes · AWS

Experience

July 2028Present

Research Scientist

OpenAI

Improved model reliability and reduced hallucination rates by 22% through a suite of evaluation frameworks and production integrations. Achieved this by designing evaluation pipelines that embedded safety checks, partnering with product, safety, and infrastructure teams to deploy models in production, and mentoring researchers on large-scale experimentation and AI alignment practices.

Summer 2026Summer 2026

AI Research Intern

Google DeepMind

Enabled large-scale evaluation of generative AI by building pipelines that processed millions of samples daily to validate retrieval-augmented generation and long-context reasoning. Delivered reproducible experimental infrastructure and presented results to senior research and engineering stakeholders to inform model direction.

Aug 2023May 2028

Graduate Research Assistant

Columbia AI Research Lab (Columbia University)

Advanced multimodal document understanding and knowledge retrieval research, resulting in publications at NeurIPS and ACL. Supported this by designing distributed training pipelines for large-scale transformer models using PyTorch and Kubernetes and collaborating on human–AI interaction studies with faculty and peers.

Education

2023May 2028

Ph.D. in Computer Science

Columbia University

Research Focus: Generative AI, Multimodal Learning, Human-AI Interaction

2018May 2022

B.S. in Electrical Engineering and Computer Science

Massachusetts Institute of Technology (MIT)

GPA: 4.8/5.0

Hosted on Silpi
John Doe | Silpi