Argho Sarkar, PhD

Senior Generative-AI Scientist @ Cotiviti

About

❖ I am a Generative AI Scientist specializing in LLMs, agentic workflows, and multimodal systems, focused on building production-grade AI platforms that deliver measurable business impact.✦ At Cotiviti, I lead the development of enterprise GenAI platforms that automate complex workflows and drive multimillion-dollar revenue impact 💰. I am designing and implementing LLM- and agentic AI–driven systems that are transforming organizations from manual, rule-based operations into dynamic, reasoning-driven decision frameworks. My work spans prompt and context engineering, automated code generation, large-scale data standardization, intelligent document and chart review, Retrieval-Augmented Generation (RAG), and LLM fine-tuning for downstream enterprise applications.✦ Previously, at Memorial Sloan Kettering Cancer Center, I built foundation-model–based cancer detection pipelines from whole-slide images, enabling more scalable and efficient diagnostic interpretation. I also leveraged LLMs to automate structured label generation from unstructured clinical reports.✦ During my internship at AWS, I developed multimodal AI systems integrating temporal satellite imagery with natural language queries to deliver high-precision, context-aware building-level damage assessments — addressing the challenge of extracting actionable insights from large-scale disaster data.✦ In collaboration with national labs, I developed object detection systems designed to reduce the performance gap between models trained on synthetic data and deployed on real-world data. By improving cross-domain generalization, I enhanced model reliability in scenarios where large-scale labeled real-world data was unavailable — enabling practical deployment in data-constrained, high-stakes applications. ⌘ Overall, my expertise lies in building scalable Generative AI and multimodal systems that bridge cutting-edge research with real-world deployment. I lead high-impact initiatives that transform complex data into intelligent, decision-ready systems, generating multimillion-dollar value and advancing applied AI. I am driven to build trustworthy, high-performance AI systems that solve complex, high-stakes problems and create impact at scale.

Country

United States

City

Washington DC-Baltimore Area

Industry

Hospital & Health Care

Skill

Agentic Workflows, Synthetic Data Generation, Object Detection, Convolutional Neural Networks (CNN), Long Short-term Memory (LSTM), BERT (Language Model), Distributed Training, Medical Imaging, High Performance Computing (HPC), Transformer Models, Leadership, Large Language Models (LLM), Generative AI Tools, Google Cloud Platform (GCP), vertex AI, Large Scale Deployments, Cross-functional Team Leadership, Large-scale Data Analysis, Explainable AI, Natural Disaster

Experience

Senior Generative-AI Scientist

Cotiviti

2026-3 - Present · 7 mos

Generative-AI Scientist

Cotiviti

2025-2 - Present · 1 yr 8 mos

AI Researcher

Memorial Sloan Kettering Cancer Center

2024-4 - 2025-2 · 11 mos

New York City Metropolitan Area

•Leading the deployment of scalable deep learning workflows for pathology Whole Slide Images, focusing on evaluating and developing foundation models to enable end-to-end cancer diagnosis and streamline the screening process. •Implementing Large Language Models (LLMs) (GPT, Llama), including Retrieval-Augmented Generation (RAG) and LangChain, to efficiently extract medical data from databases using natural language queries. •Skills: Health AI, HPC, PyTorch, Multiple Instance Learning, Representation Learning, CLIP, Vision Transformer, Fine-tuning, LLMs

Graduate Research Assistant (GRA)

University of Maryland Baltimore County

2019-7 - 2024-4 · 4 yrs 10 mos

Baltimore, Maryland Area

•Implemented a vision transformer to extract spatiotemporal features from sequential aerial video for anomaly detection. •Developed large-scale VQA datasets (FloodNet & RescueNet) for high-resolution remote-sensing images. •Developed an explainable Visual QA framework for large-scale remote sensing datasets by integrating Grad-CAM-based auxiliary visual supervision, which enhanced model accuracy, improved visual explanations, and reduced manual annotation efforts. •Enhanced model robustness and consistency against adversarial manipulation and image rotation for image classification tasks by proposing consistency and robustness loss, achieving a 3% improvement in accuracy and explanations on benchmark datasets. •Designed and deployed TinyVQA for resource-constrained devices using knowledge distillation, achieving a 100% memory reduction with only a 1.5% accuracy loss, enabling real-time operations on drones with a power requirement of just 0.7 W. •Skills: PyTorch, Multi-GPU Training, Multimodal AI, Edge AI, CNN, LSTM, BERT, Attention, Explainability, Knowledge Distillation

FDL Summer Research Fellow

SETI Institute

2023-6 - 2023-8 · 3 mos

Mountain View, California, United States

•Optimized the synthetic image-based object detection workflow by ensembling YOLO-V8 with Vision Transformer (ViT) and Masked Auto-Encoder models, achieving a 30% improvement in synthetic-to-real performance to enhance nuclear safeguards. •Collaborated with Sandia National Lab, DOE, NVIDIA, and Google Cloud. •Skills: Python, PyTorch, GCP, CNN, Object Detection, Transformer, Masked Auto-Encoder, Multidisciplinary Collaboration, Leadership

Applied Scientist Intern

Amazon Web Services (AWS)

2022-5 - 2022-9 · 5 mos

Virginia, United States

•Developed a machine learning pipeline and a novel dataset that combines pre- and post-disaster satellite imagery with a question-answering framework, enabling precise and context-aware building-level damage assessments for first responders. •Designed a transformer-based (self-attention) deep multimodal fusion framework, resulting in a 10% improvement in damage assessment accuracy over single-image-based (post-disaster) approaches.

Statistician

Child Health Research Foundation (CHRF)

2018-8 - 2019-1 · 6 mos

Bangladesh

•Data management for health survey data using SQL. •Exploratory data analysis on large-scale, high-dimensional health survey data. •Statistical data analysis: data cleaning, feature selection, hypothesis testing, predictive modeling.