Prathyusha Akundi

Hello, I'm

Prathyusha Akundi


As a Senior Data Scientist, I design end-to-end production systems that handle underwriting, fraud detection, and clinical intelligence at scale—turning complex problems into efficient, automated solutions.

/ Career

A brief overview of my career

Mar 2021 - Present

Senior Data Scientist

Perfios Software Solutions Pvt. Ltd.

Senior Data Scientist designing end-to-end agentic workflows focused on FinTech and Healthcare use cases. Experienced in LLM fine-tuning, agent orchestration, knowledge graph architecture, reinforcement learning, and evaluation frameworks—with growing focus on AI safety paradigms.

GenAIComputer VisionMachine LearningDeep LearningPythonSQLReinforcement Learning
Aug 2016 - Jun 2018

Business Technology Analyst

Deloitte USI

Delivered insurance technology projects on the Guidewire suite (Policy Center, HPExstream, GOSU); 3 Applause and 4 Spot awards.

GuidewireGOSUHPExstreamPolicy Center
Jun 2015 - Aug 2015

Intern - Software Engineer

Microsoft Corporation

Enhanced Course Publishing analytics using SQL. Developed 'ShareIt', a file sharing solution with Azure integration.

SQLAzure

/ Publications

Research papers, workshop contributions, and academic writing.

International Conference on Document Analysis and Recognition · 2024

Reformulating Key-Information Extraction as Next Sentence Prediction for Hierarchical Data

Reformulating Key-Information Extraction for hierarchical tables as Next-Sentence Prediction to improve information extraction from complex documents.

International Symposium on Biomedical Imaging (ISBI) · 2022

Incremental Learning for a Flexible CAD System Design

Mitigating Catastrophic Forgetting through auxiliary task representations to enable incremental class learning in medical imaging systems.

Indian Conference on Computer Vision, Graphics and Image Processing · 2021

Manifold Learning to address Catastrophic Forgetting

Addressing Catastrophic Forgetting in medical CAD systems through manifold learning, enabling adaptation across sequential datasets from different imaging sources.

arXiv · 2021

A Clinically Inspired Approach for Melanoma classification

Improving melanoma detection in CAD systems through Intra-Patient Comparative Analysis to identify outlier lesions and boost diagnostic accuracy.

/ Tech Stack

Breakdown of my technical skills, categorized by domain and proficiency

Expert
Proficient
Beginner

Core AI & Machine Learning

Machine Learning
Deep Learning
Computer Vision
NLP
Image Processing

Generative AI & Agents

Agentic AI
LLM Fine-tuning
Reinforcement Learning
Evals & Alignment
Knowledge Graphs

Languages & Databases

Python
SQL
Cypher (Neo4j)

Tools & Frameworks

FastAPI
Docker
Streamlit
LangGraph
ADK
Smol Agents

/ Contact Me

Feel free to reach out. I'm always open to discussing new projects and opportunities.