Professional Summary
Overview
Jothsna Praveena Pendyala is an AI Platform Architect and Senior Data Scientist at Infosys, serving as the principal architect of an enterprise Agentic AI platform at AT&T — one of the largest telecommunications companies in the United States. She architects and owns a production-scale, multi-agent AI system deployed across AT&T's engineering organization, built on LangGraph, Azure Kubernetes Service, and a custom MCP server infrastructure.
Her expertise spans agentic AI systems design, Kubernetes platform engineering, LLM infrastructure, and AI observability at enterprise scale. She has prior experience on Amazon's AGI team, working directly on the Nova family of large language models. She is a published researcher with peer-reviewed work in IEEE, Springer, and other indexed venues, an editor of a forthcoming book published by Taylor & Francis, and a contributing author to four books in the field of AI and data engineering.
She is an IEEE Senior Member, holds elected and appointed leadership roles at the Association for Computing Machinery (ACM) — including Chair and Chief Editor of the ACM Dallas Chapter — and has been an invited panelist at major international technology conferences including the Linux Foundation's Open Source Summit North America 2026.
Career
Professional Experience
- Architected and owns an enterprise Agentic AI Platform-as-a-Service (PaaS) enabling reusable multi-agent workflows across AT&T's engineering organization
- Designed full end-to-end system architecture: LangGraph orchestration, MCP-based tool integration, microservices backend, and Azure Kubernetes Service (AKS) deployment
- Serves as product owner for the AI platform — defines requirements, prioritizes the roadmap, and drives adoption across multiple business teams
- Built and manages MCP (Model Context Protocol) servers enabling seamless integration of enterprise tools, APIs, and data sources within a unified agentic platform
- Developed scalable backend services using Python and FastAPI; deployed on AKS with Kubernetes autoscaling (HPA/KEDA), supporting high-concurrency, low-latency production workloads
- Established full observability stack using LangSmith, Arize, and OpenTelemetry for tracing, debugging, and performance monitoring of LLM-powered systems
- Implemented LLM guardrails (LiteLLM) and Veracode SAST/SCA security scanning for compliant, secure AI deployment
- Manages CI/CD pipelines via Azure DevOps for automated deployments and consistent release management across environments
- Directs a team of 7–8 senior data scientists; conducts daily pull request reviews and quarterly architecture reviews with authority to approve, request changes, or reject proposals
- Leads stakeholder collaboration across product, engineering, and business units; drives platform adoption across the enterprise
- Built intelligent agents on Infosys Agentic Foundry to automate complex data mapping and consolidation from disparate spreadsheet and PDF sources into structured formats
- Created and annotated over 25,000 high-quality multimodal training and evaluation examples (text, image, audio, video) for Amazon Nova large language models
- Collaborated directly with Amazon scientists to design guardrail strategies — prompt rephrasing, output constraints, and filtering — contributing to a measurable 18% improvement in factual accuracy
- Maintained a personal annotation accuracy score of 98%+ across all tasks
- Documented 50+ edge case resolution strategies to standardize handling of complex annotation scenarios
- Filed 30+ actionable UI/UX improvement tickets; 3 major platform updates were implemented as a result
- Designed and deployed a conversational AI agent using GPT-4o, AWS Bedrock, and foundation models including Claude, Mistral, and LLaMA
- Engineered Retrieval-Augmented Generation (RAG) pipelines for customer tariff data retrieval; achieved 95% accuracy improvement in billing assessments
- Fine-tuned GPT-4o on utility data and FAQs; delivered 20% improvement in response accuracy across 3,000+ interactions
- Built ETL pipelines using AWS Glue and Redshift for model retraining; reduced manual preprocessing time by 25%
- Developed and deployed ML models (XGBoost, Random Forest) for short-term energy demand forecasting; achieved 89% accuracy across 100,000+ records
- Monitored and maintained electrical infrastructure across 200+ transformers and 4 substations using IoT-based predictive maintenance
- Designed regulatory compliance reporting workflows ensuring 100% adherence to government-mandated operational performance standards
- Taught engineering subjects, mentored students, and guided academic projects, building a foundation in technical teaching and mentorship
Academic Background
Education & Certifications
Education
Certifications
- Claude Certified AI Architect Anthropic
- GitHub Copilot Certified Engineer GitHub / Microsoft
- AWS Certified AI/ML Professional Amazon Web Services
Scholarly Output
Research Publications
Full list with citations: Google Scholar Profile
Book Contributions
Editor — RAG in Action
Editor responsible for overseeing the compilation, technical review, and editorial direction of this practitioner-focused book on Retrieval-Augmented Generation systems, including proposal ownership, author selection, and review of 11 chapters.
Contributing Author — Book Chapters
- "Building a Generative AI Application: A Case Study on an AI Travel Planner," in Generative AI in Software Development, pp. 289–323, 2026
- Two forthcoming chapters, Artificial Intelligence Agents in Industrial Orchestration, Elsevier
Recognition
Awards & Honors
- Best Paper Award — DASGRI 2026, for "Multimodal Data Analytics: Advanced Fusion Architectures for Cross-Modal Sentiment Analysis and Customer Behavior Understanding," School of Computing, Goldsmiths, University of London, published by Springer (April 10–11, 2026). Conference media coverage
- Invited Panelist, Linux Foundation Open Source Summit + Embedded Linux Conference North America 2026, representing Infosys on production-ready AI agents
- Infosys InstaRise Award (special category), for exceptional performance, client engagement, and rapid implementation of proofs of concept
- Starshine Women Award, recognition through Infosys Women in Tech Series
Community & Service
Professional Memberships & Leadership
The highest grade of membership attainable through application, requiring a minimum of 10 years of professional experience and demonstrated significant performance in IEEE fields of interest. Fewer than 10% of IEEE's 400,000+ members hold Senior Member grade.
The world's largest scientific and educational computing society, with over 100,000 members globally. Holds multiple elected and appointed leadership roles.
Judging, Reviewing & Editorial Service
- Peer reviewer for 30+ scientific publications submitted to IEEE and Scopus-indexed journals and international conferences across AI, ML, data science, NLP, agentic AI, and computer vision
- Primary editor of the forthcoming RAG in Action book — proposal ownership, author selection, review of 11 chapters, and manuscript submission to Taylor & Francis
- Quarterly architecture reviewer for enterprise Agentic AI platform features and services proposed by product owners across multiple teams
- Daily code reviewer for a team of 7–8 senior data scientists on production platform work
- Final-decision panelist for technical hiring interviews at Infosys for data science and AI engineering roles
- Speaker proposal reviewer for Society of Women Engineers and ACM Dallas Chapter programming
Public Engagement
Speaking & Media Presence
Evals: One Does Not Simply Ship an Agent
Plano Prompt Engineers · Improving, Plano, TX · July 22, 2026
Hands-on workshop on evaluation frameworks for AI agents — building a working eval system around a Claude-powered shopping assistant, covering task definitions, deterministic and LLM-as-judge grader types, and iterative performance improvement strategies.
Presenting at Plano Prompt Engineers, July 2026
Scaling AI Workloads on K8s: Reliability Patterns for Prod. Agentic Platforms
Kubernetes and Cloud Native North Dallas · Improving, Plano, TX · July 16, 2026
Presented on production challenges in AI agent deployment, drawing on enterprise experience operating agentic platforms on Azure Kubernetes Service — designing for horizontal scalability across heterogeneous AI workloads and reliability patterns for cluster management.
Presenting at Kubernetes and Cloud Native North Dallas, July 2026
Open Source Summit + Embedded Linux Conference North America 2026
Linux Foundation · May 19, 2026
Invited panelist for "Building an Enterprise Platform for Production-Ready AI Agents." Represented Infosys alongside panelists from SAS, DeployHub, and Southwest Airlines — the panel brought together enterprise AI platform leaders to discuss production-grade agentic AI deployment at scale.
Panel discussion, Open Source Summit North America 2026
Instagram — @jothsna.aitales: AI education and practitioner content reaching a growing audience of engineers and data scientists
Newsletter — The Agentic Stack: practitioner-focused publication covering production agentic AI, Kubernetes platform engineering, and LLM infrastructure
Open Source
GitHub Contributions
Code, tooling, and open-source activity on GitHub — including MCP server integrations, agentic AI utilities, and data engineering projects.
View GitHub Profile →Capabilities