Soundrya Kesari

AI/ML professional with hands-on experience building Generative AI and Agentic AI systems for enterprise use cases. Designs and evaluates AI architectures, develops Retrieval-Augmented Generation (RAG) pipelines, and orchestrates multi-agent workflows using LangGraph and LLMs. Combines strong Python, NLP, and cloud skills (Azure, GCP) to turn business needs into scalable, production-ready solutions. Researches emerging AI technologies and recommends practical, compliance-aware approaches that drive measurable innovation.

0
Projects
0
Skills
Work Experience
AI/ML Computational Science Analyst
Accenture Solutions Pvt. Ltd.
Jun 2026 — PresentCurrentGurugram, India
  • Engineered multi-agent AI workflows with LangGraph, automating research, document analysis, and structured report generation for enterprise teams
  • Constructed RAG pipelines integrating ChromaDB and Azure OpenAI to deliver contextual answers from internal knowledge bases
  • Assessed Agentic AI frameworks—focusing on orchestration patterns and tool-use strategies—and endorsed architectures that comply with governance standards
  • Authored comprehensive technical guides and architecture blueprints, empowering project teams to consistently implement LLM-powered solutions across the organization
AI/ML Computational Science Associate
Accenture Solutions Pvt. Ltd.
Jul 2025 — May 2026Gurugram, India
  • Developed and tested RAG-based systems using Azure AI Search and GCP Vertex AI, improving response accuracy for internal Q&A tools by embedding enterprise data sources
  • Ran proof-of-concept experiments for Agentic AI use cases—prioritizing autonomy, safety, and domain-specific reasoning—and presented findings to AI leadership
  • Benchmarked emerging frameworks (LangChain, LlamaIndex, Autogen) and documented trade-offs in latency, cost, and scalability for each
  • Collaborated with cloud architects to deploy LLM services on Azure and GCP, ensuring secure API access and cost-efficient scaling
Associate Software Engineer
Accenture Solutions Pvt. Ltd.
Jul 2024 — Jun 2025Gurugram, India
  • Scripted Python automation for data preprocessing and model inference pipelines, reducing manual effort in AI-driven workflows by 30%
  • Contributed to Agile delivery cycles—writing unit tests, debugging model outputs, and integrating ML modules into client-facing applications
  • Partnered with data scientists and product owners to refine feature requirements for NLP and computer vision components
Software Trainee
Persistent Systems
Jan 2024 — Jun 2024Pune, India
  • Built REST APIs and backend services in Java using Spring Framework, following object-oriented design principles
  • Participated in code reviews and version control workflows, catching performance bottlenecks before deployment
  • Developed unit tests and integration tests, maintaining >90% code coverage for assigned modules in a client-facing project
Projects
Multi-Agent Enterprise Research Assistant

Designed LangGraph agent workflows (Planner, Research, RAG, Analysis, Report) for multi-step reasoning and structured report generation on complex topics. Implemented Retrieval-Augmented Generation (RAG) with ChromaDB vector embeddings to ground LLM responses in enterprise knowledge sources. Built FastAPI backend services and a modular agent architecture that supports scalable deployment and easy integration with Azure OpenAI. Automated citation-backed summarization and source tracking, delivering actionable recommendations alongside each generated report.

PythonLangGraphChromaDBFastAPIAzure OpenAI
Face Detection and Recognition System

Built a real-time facial recognition pipeline using OpenCV and deep learning, cutting average verification time from 10 seconds to under 3 seconds. Implemented face detection, feature extraction, and identity verification components with optimized image processing for better throughput. Improved recognition accuracy by tuning model parameters and preprocessing steps, enabling reliable performance across varied lighting conditions.

PythonOpenCVDeep Learning
Text Classification using Named Entity Recognition

Developed a text classification system that uses Named Entity Recognition (NER) for entity extraction and feature engineering, achieving over 90% accuracy. Designed an NLP pipeline that reduced document processing time by approximately 30% while improving automated categorization and information extraction. Enhanced downstream search and analytics capabilities by integrating extracted entities directly into the classification workflow.

PythonNLPMachine Learning
Skills
Large Language Models (LLMs)Agentic AILangGraphRAG (Retrieval-Augmented Generation)Prompt EngineeringAI WorkflowsMulti-Agent SystemsPythonJavaSQLMachine LearningDeep LearningNatural Language Processing (NLP)Computer VisionPyTorchTensorFlowKerasMicrosoft AzureGoogle Cloud Platform (GCP)Azure OpenAIGitLinuxREST APIsChromaDBFastAPIAgile DevelopmentSDLC
Education
Bachelor of Technology, Computer Science and Engineering
Noida Institute Of Engineering and Technology
Nov 2020 — May 20248.8/10
Extracurricular Activities
National Science Olympiad – Gold Medalist

Earned Gold Medal for academic excellence in science and mathematics, demonstrating strong analytical and problem-solving abilities.

International Dance Competition (Ras Banaras) – First Runner-Up

Placed second in an international classical dance competition, requiring discipline, precision, and performance under pressure.

Classical Dance Training – Kathak & Bharatanatyam

Completed formal training in Indian classical dance forms, building focus and dedication through years of practice.