Generative AI Engineer · Tempe, AZ

I build the
agents that reason,
retrieve & ship.

I design LLM-powered applications — document QA, code generation, and multi-agent workflows — with LangChain, LangGraph, and AutoGen. Grounded in deep learning, prompt engineering, and production software in Python and C++.

Get in touch → GitHub LinkedIn
CurrentlyCognizant
FocusMulti-Agent AI
StackLangChain · GPT-4
01 Profile

I'm a Generative AI Engineer who turns large language models into systems people can actually use — retrieval pipelines that answer questions over real documents, assistants that route tasks intelligently, and agent teams that collaborate across the software development lifecycle.

Before generative AI, I spent three years at Philips shipping C++ tooling and machine learning pipelines in medical imaging, where a GAN I built cut patient X-ray radiation exposure in half. That mix — rigorous engineering plus applied ML — is how I approach every problem.

I hold a Master of Science in Computer Software Engineering from Arizona State University, with earlier training in AI/ML from IIIT Bangalore and a B.Tech in Computer Science from PES University.

Languages & Frameworks

C++PythonSQLC LangChainLangGraphAutoGenChainlit TensorFlowKerasPyTorchAzure

AI / ML

Generative AIDeep LearningNLP GANsSocial Network Analysis

Tools & Platforms

PostgreSQLOracleSQLMySQL GitGitHubJenkinsLinux/Unix
02 Experience
Generative AI Engineer — Associate
Dec 2024 — Present
Cognizant · Tempe, AZ
  • Built a document Q&A system with LangChain, GPT-4 and FAISS that ingests PDF/DOCX, chunks and embeds content, and answers questions via OpenAI function calling.
  • Developed a LangChain assistant that classifies queries into code-generation or general tasks and routes them through dynamic prompt templates for tailored responses.
  • Designed a multi-agent AutoGen group chat spanning requirements analysis, code generation, test generation and document Q&A — backed by ChromaDB vector memory for context-aware, long-term continuity.
Software Engineer 2
Jul 2019 — Aug 2022
Philips India Ltd. · Bangalore
  • Built a Pix2Pix GAN with imaging specialists to upscale low-dose X-rays to high-dose quality, cutting patient radiation exposure by 50%.
  • Deployed a call-classification pipeline (word2vec + Linear SVC) that reduced manual triage from 3 weeks to 30 minutes and condensed 100+ issue categories down to 15.
  • Developed a C++ diagnostic tool to pinpoint corruption in the image pipeline, validated with the GTest framework.
  • Prototyped UX and image-quality POCs, including resizable UI panels and 4K rendering at lower resolutions.
Software Engineering Intern
Jan 2019 — Jul 2019
Philips India Ltd. · Bangalore
  • Built a Python remote-connectivity package for X-ray machines, integrated with the service tool to cut machine access time by 50% and reduce on-site visits.
Data Science Intern
Jun 2018 — Aug 2018
IBM India Ltd. · Bangalore
  • Surfaced the top 10 usability issues in SBI's YONO banking app by analyzing Play Store reviews with Watson Natural Language Understanding.
03 Selected Work
Master's Thesis · 2023–2024

Thoracic Pathology Classification & Localization

A deep learning framework combining Swin Transformers, custom CNNs and ensemble methods to classify 15 thoracic diseases and localize abnormalities on the NIH Chest X-ray dataset.

avg AUC 0.60 · 15 disease classes · spatial localization
Production · Cognizant

SDLC Multi-Agent Assistant

An AutoGen group-chat system where specialized agents handle requirements, code, tests and document Q&A, with ChromaDB-backed persistent memory for context-aware retrieval.

AutoGen · ChromaDB · GPT-4 · vector memory
Production · Cognizant

Document Q&A over PDF / DOCX

A retrieval pipeline that chunks and embeds uploaded documents, then answers natural-language questions through OpenAI function calling.

LangChain · GPT-4 · FAISS embeddings
Medical Imaging · Philips

Low-Dose X-ray Enhancement GAN

A Pix2Pix GAN translating low-dosage X-rays into high-dosage quality, reducing the need for repeat scans and lowering patient radiation exposure.

−50% radiation exposure · Pix2Pix · clinical collaboration
04 Publication

Predicting Protein-Protein Interaction in Multi-layer Blood Cell PPI Networks

Springer · Advanced Informatics for Computing Research · Jun 2019
View DOI →

Let's build something that thinks.

Open to conversations about generative AI, multi-agent systems, and applied machine learning.

skhare10@asu.edu LinkedIn GitHub