Data Engineer • GenAI Platforms • Chennai

Hey, I'm Lokeshwar V

Data pipelines that move the data, multi-agent systems that answer for it.

Two and a half years at Accenture building enterprise ETL pipelines, then the multi-agent AI platform that sits on top of them — CrewAI orchestration, RAG over Milvus, and citation-backed ChatOps answers delivered straight into Microsoft Teams.

Hire MeDownload CV
Incident triage: 30–45 min → under 5
Dashboards: 6 min → 2 min (~67%)
1,100+ traceable reports in 30 days

Orchestration flow

  1. Extractsource systems
  2. TransformPandas
  3. Embed
  4. IngestMilvus + SQL
  5. Queryfrom Teams
  6. Intent Router
  7. AgentsCrewAI
  8. Cited Answer

About Me

I am a Data Engineer at Accenture who builds two things that meet in the middle: the ETL pipelines that get enterprise data into SQL and Milvus, and the multi-agent AI platform that lets an engineer ask a question about that data in Microsoft Teams and get a citation-backed answer seconds later.

The interesting part is usually the reliability work — hallucination guardrails, automated citation validation, retry logic for LLM rate limits — the things that decide whether an agent is a demo or something a team actually leans on in production.

500K+ Records/Run5 Microservice Environments20 Visualization TypesStar of the Month 2025

Core stack

ETLPythonSQLMilvusRAGCrewAIAzure OpenAIFastAPILogic Apps

Professional Experience

Mar 2026 – Present

Packaged Application Development Analyst

Accenture India, Chennai

  • Architected and built an enterprise multi-agent AI orchestration platform on CrewAI, FastAPI, Azure OpenAI (GPT-4o / o3), Anthropic Claude and Milvus, automating intelligence workflows and ChatOps operations.
  • Cut manual DevOps troubleshooting from 30–45 minutes to under 5 minutes per incident.
  • Hardened the platform with hallucination guardrails, automated citation validation, exponential-backoff retries for LLM rate limits and TLS-secured vector connectivity — from frequent hallucinated responses to virtually none in production.
  • Engineered dynamic intent classification and multi-path routing across specialized autonomous agents, with real-time ChatOps querying through Microsoft Teams Adaptive Cards across 5 isolated microservice environments.
  • Built an asynchronous dashboard and infographic engine on a two-agent Strategy/Rendering pipeline covering 20 visualization types, cutting generation time from 6 minutes to 2 (~67%).
  • Added source traceability to that engine, powering 1,100+ traceable reports in its first 30 days.
  • Design and manage the scalable data pipelines behind enterprise dataset ingestion and retrieval in Python and SQL, and debug production failures in existing retrieval pipelines.

Apr 2024 – Mar 2026

Packaged Application Development Associate

Accenture India, Chennai

  • Designed and implemented automated ETL pipelines for large enterprise datasets using Python and Pandas.
  • Built scalable ingestion workflows to extract, transform and load into SQL databases and the Milvus vector database.
  • Developed Microsoft Teams event-based automation where keyword triggers activate Azure Logic Apps to start backend data operations.
  • Integrated Logic Apps with APIs hosted on Azure App Service to orchestrate data loading, pipeline execution and job monitoring.
  • Processed datasets exceeding 500K records with validation and transformation logic for data quality assurance.
  • Deployed Python backend services on Azure App Service supporting internal automation systems and data workflows.

Dec 2023 – Apr 2024

Software Engineer Intern

NYL Technology, Chennai

  • Contributed to backend API development and data processing tasks using Python and REST APIs.
  • Worked on early LLM retrieval prototypes with LangChain, Pinecone and Flask.

Featured Projects

  • Confidential Client Work
    2026Accenture, India
    Multi-Agent Dashboard & Infographic Engine

    An asynchronous two-agent pipeline: a Strategy agent decides what the chart should say, a Rendering agent draws it. Covers 20 visualization types and traces every figure back to the record it came from.

    Generation time down from 6 minutes to 2 (~67%); 1,100+ traceable reports in the first 30 days.

    Flow:Request -> Strategy Agent -> Rendering Agent -> Traceable Report

    PythonFastAPICrewAIAsync Pipelines
  • Confidential Client Work
    Mar 2026 - PresentAccenture, India
    Enterprise Multi-Agent AI Orchestration Platform

    An intent classifier routes a Teams question down one of several paths, each handled by an autonomous agent that pulls from SQL and Milvus before answering. Guardrails and automated citation validation sit between the model and the reply, so every answer points back at its source.

    DevOps triage cut from 30-45 minutes to under 5, across 5 isolated microservice environments.

    Flow:Teams Query -> Intent Router -> Specialized Agents -> Milvus + SQL -> Guardrails -> Cited Answer

    CrewAIFastAPIAzure OpenAI GPT-4o / o3Anthropic ClaudeMilvusTeams Adaptive Cards
  • Confidential Client Work
    Nov 2025 - PresentAccenture, India
    Customized AI Agent with Milvus & SQL Retrieval

    Built a prompt-driven AI agent that performs tool-calling across Milvus vector search and SQL retrieval, then returns citation-backed answers for reliable enterprise usage.

    Flow:Extract -> Store -> Transform -> Embed -> Ingest -> Prompt -> Tool Calling -> Retrieval -> Response with Citations

    PythonMilvusSQLSemantic SearchAI Agents
  • Confidential Client Work
    Nov 2025Accenture, India
    Automated GenAI Document Processing Pipeline

    Automated the full ingestion flow from Blob Storage to Milvus with PDF conversion, GPT summarization, chunking, embeddings, metadata enrichment, and daily delta synchronization.

    Flow:Extract -> Store -> Transform -> Embed -> Ingest -> Prompt -> Tool Calling -> Retrieval -> Response with Citations

    PythonAzure Blob StorageGPTMilvusVector Embeddings
  • Confidential Client Work
    Apr 2025Accenture, India
    Azure DevOps Workflow Automation Engineer

    Automated Azure DevOps taskboard lifecycle with daily task creation, stale-task closure, monthly story initialization, and weekly recap emails using Python plus Logic Apps.

    PythonAzure DevOps APIAzure Logic AppsWorkflow Automation
  • Confidential Client Work
    Oct 2024 - Oct 2025Accenture, India
    Data Automation Engineer | Python Developer | Vector Database Integration

    Developed an end-to-end Python automation pipeline that ingests high-volume datasets, transforms them with pandas, and inserts vectorized outputs into Milvus for fast similarity retrieval.

    Flow:Extract -> Transform -> Embed -> Ingest

    PythonpandasMilvusCron Jobs
  • Confidential Client Work
    Mar 2024 - Apr 2024NYL Technology
    AI Assistant for Manual and Automation Testing

    Designed an assistant-oriented workflow to support test case preparation, execution support, and faster QA feedback loops for manual and automation teams.

    PythonNLPPrompt EngineeringQA Automation
  • Confidential Client Work
    Feb 2024NYL Technology
    GMeet Summarizer

    Created a meeting intelligence flow where Google Meet transcriptions stream through Pub/Sub and are summarized by ChatGPT for concise post-meeting outcomes.

    Google WorkspaceGoogle MeetGoogle Pub/SubChatGPT

Project Archive

Education & Certifications

2023 – 2025

Master of Business Administration — Data Science

Amity University · CGPA 7.4

2018 – 2022

Bachelor of Technology — Information Technology

Vel Tech Multi Tech Engineering College · CGPA 8.15

Azure AI Engineer Associate (AI-102)

Microsoft Certified

Azure AI Fundamentals

Microsoft Certified

Accelerate AI-assisted development using GitHub Copilot

Microsoft Applied Skills

Software Industry Foundations

Primerli

Impact & Achievements

2.5 Years

Building data & AI systems

500K+

Records processed per pipeline

< 5 min

Incident triage, down from 30–45 min

1,100+

Traceable reports in first 30 days

HIGH FLYER — Star of the Month, Accenture (2025)
TechExpressway Merit Holder, Accenture (2024)
TCS National Qualifier Test — 84%
Dashboard generation cut ~67%, from 6 minutes to 2
ChatOps rolled out across 5 isolated microservice environments
Hallucinated responses reduced to virtually none in production

Let's Connect

I am actively looking for Data Engineer roles and impactful AI/data platform projects. If you have an opportunity or collaboration idea, let's talk.

Email: lokeshwar_v@yahoo.com

Location: Chennai, India

GitHubLinkedIn