Varun Mannam

About Varun Mannam

Personal Website: https://varunmannam.github.io/I like solving unknown things in smarter ways.(AI/ML research in Generative AI/ LLMs/NLP/NLU/computer vision)Passionate in machine learning, specifically in deep learning algorithms for text, image and signal data processing. Currently, I am working on automation for the high-quality data generation using deep learing methods for text, image and video modalities in Alexa AI, Amazon. I worked on developing state of art deep learning models for image denoising, super-resolution, and compressive sensing of biomedical images (captured by microscopy) during my graduate studies at the University of Notre Dame.For more details about my publications, please go to my Google Scholar & hl=en & oi=ao

Experience

1. Applied Scientist Ii

   Amazon Web Services (AWS)

   Jul 2025 — Present · Seattle, WA, US

   Working with Dr. Chaitanya S in the AWS Healthcare AI team.I am a core science contributor to Project Pulse, an AI-powered healthcare contact center built on Amazon Connect (Q in Connect). I led end-to-end science design, prompt engineering, and evaluation of multiple LLM-based agents across milestone releases, including an Appointment Management Agent orchestrating EHR interactions (Epic APIs) for real-time scheduling; a Patient Verification Agent using LangGraph architectures for secure identity authentication; and an Insurance Verification Agent for eligibility checks. I designed evaluation frameworks encompassing goal fulfillment metrics, latency benchmarking, safety guardrails, and synthetic data generation across text and voice (STT/TTS) modalities.I evaluated the Nova Sonic voice AI model for healthcare workflows, benchmarking against alternatives for production readiness. I led the Nova Forge customization initiative, designing SFT strategies with LoRA on Nova Lite (~20B parameters) using 250+ medical notes across physical and behavioral health formats (SOAP, GIRPP, BIRP, DAP, SIRP).For AWS HealthScribe, I analyzed multi-speaker medical conversations, quantifying summarization degradation when 3+ speakers are present and identifying vulnerable clinical note sections to inform model improvement roadmaps.I submitted a patent for Smart Healthcare Scheduling with Natural Language Intelligence and authored research papers for AMLC, KDD workshops.Keywords: Healthcare AI, LLM Agents, Generative AI, RAG, Prompt Engineering, Multi-Agent Systems, Amazon Connect, LangGraph, Nova Sonic, Nova Forge, SFT/LoRA, AWS Bedrock, HealthScribe, EHR Integration, Epic APIs, Medical NLP, MLOps

Education

• University of Notre Dame

  Doctor of Philosophy - PhD, Electrical Engineering

  2017 — 2022

• Bapatla Engineering College

  Bachelor of Technology (B.Tech.), Electronics & Communication Engineering

  2006 — 2010

• Indian Institute of Technology, Kharagpur

  Master of Technology (M.Tech.), Electronics and Communications Engineering

  2010 — 2012

Skills

• Agilent Ads
• Data Structures
• Digital Electronics
• 802.11n
• Microwave Engineering
• Matlab
• Ieee Standards
• 5g
• Embedded Systems
• Phy
• Signal Processing
• 802.11a
• Systemvue
• 802.11ad
• Rf
• 802.11ax
• Lte
• Wlan
• Fpga
• Control Systems Design
• Mac Layer
• Electronics
• Ieee 802.11
• Passive Components
• Data Link Layer
• Wifi
• C++
• Cst Microwave Studio
• Bluetooth
• Wireless
• Ni Labview
• Antenna Measurements
• Cellular Standards
• Connectivity
• 802.11p
• Labview
• C