HELLO, I'M

Balakumaran R

GenAI / AI Engineer

• • • •

Chennai, India | Open to Relocation

8 years across generative AI, machine learning and data engineering. Currently building production LLM and agentic AI systems at Johnson Electric.

About Me

AI Engineer with 8 years of experience across generative AI, machine learning and data engineering.

Currently at Johnson Electric, building production LLM and agentic AI systems: a PLM support chatbot, and an MCP-based AI agent on Azure OpenAI GPT-5 with a Neo4j knowledge graph serving engineering and quality users.

Strong data engineering foundation in Databricks, PySpark, Kafka and Airflow. Mentors junior engineers and works directly with business stakeholders.

Professional Experience

Data Science Engineer

Johnson Electric

Jan 2024 - Present · Chennai (Hybrid)

  • Mentor a team of 3 junior engineers; present solutions to stakeholders and work directly with business users.
  • Built and run the PLM Support Chatbot and Product Knowledge AI Agent (see Featured Projects).
  • Built predictive models on time-series test data.

Career Break

Aug 2023 - Jan 2024 · Health and well-being

Software Engineer

Resulticks

May 2023 - Aug 2023 · Chennai

  • Built ETL pipelines on Google Cloud Platform, loading and processing data in BigQuery.

Data Engineer

bitsCrunch (Germany)

Jun 2022 - Apr 2023 · Remote from India

  • Built real-time Kafka streaming pipelines ingesting multi-chain blockchain data (NFTs, digital assets) into Azure Blob Storage.
  • Cleaned and preprocessed blockchain data with PySpark on Azure Databricks.
  • Maintained Airflow pipelines supporting the company's NFT analytics platform.

Data Engineer

OptiSol Business Solutions

Sep 2021 - Jun 2022 · Chennai

Nutrien (Australia) enterprise data lake.

  • Built Spark pipelines on Azure Databricks migrating data from SAP, Plasma, FMIS, Rewards, PostgreSQL and AWS S3 into Delta Lake tables.
  • Built data dictionaries.

Machine Learning Engineer

OptiSol Business Solutions

Aug 2019 - Sep 2021 · Chennai

  • MachineVantage (USA): Flask REST APIs on AWS with Jenkins, ~1 TB of MongoDB data, Airflow pipelines and API monitoring.
  • MachineVantage (USA): Azure Data Factory ETL and OLAP cubes with Power BI.
  • E-Watercooler (USA): Rasa NLU chatbot in Microsoft Teams using Microsoft Graph API to schedule group calls; Docker, Jenkins, Azure Key Vault.

Software Engineer

Trumatics Software Labs

Jun 2018 - Jul 2019 · Chennai

  • Decision tree price prediction model in Python, served via Flask API on Linux and Heroku.
  • Selenium API testing.

Technical Skills

GenAI & LLMs

Azure OpenAI (GPT-5, GPT-4, GPT-4o mini)
Azure AI Foundry
RAG
LLM-based retrieval
Agentic AI
LangChain
LangGraph
MCP
FastMCP
Tiny Agents

LLM Apps & Observability

Open WebUI
Chainlit
Streamlit
Arize Phoenix
Grafana

Machine Learning

Time-series modeling
LSTM
Kalman filter
NLP
Rasa NLU
TensorFlow
scikit-learn
Pandas
NumPy

Data Engineering

Azure Databricks
PySpark
Delta Lake
Kafka
Airflow
Azure Data Factory
Snowflake
AWS Glue
BigQuery

Databases

Neo4j
Marqo
ChromaDB
MongoDB
PostgreSQL
MySQL
MS SQL Server

Languages & APIs

Python
SQL
FastAPI
Flask
REST APIs
Microsoft Graph API

Cloud & DevOps

Azure
AWS
GCP
Docker
Jenkins
CI/CD
Linux

BI

Power BI

Featured Projects

Johnson Electric

Agentic AI · Production

Product Knowledge AI Agent

Agentic chatbot on Azure OpenAI GPT-5 for engineering and quality users to query product information on motors and actuators.

  • Product data modeled in a Neo4j knowledge graph.
  • Agent functions exposed as FastMCP (Model Context Protocol) tools, orchestrated with Tiny Agents.
  • LLM observability with Arize Phoenix; Grafana dashboards for usage, token cost and token expiry.
Azure OpenAI GPT-5Neo4jFastMCPTiny AgentsArize PhoenixGrafana
GenAI · Production

PLM Support Chatbot

End-to-end chatbot over SharePoint pages and wiki docs extracted via Microsoft Graph API.

  • Started as RAG with Marqo vector DB, Azure OpenAI, LangGraph and FastAPI.
  • Redesigned to LLM-based retrieval to improve accuracy: GPT-4o mini selects relevant JSON chunks, a GPT-4 agent answers.
  • Open WebUI front end with in-chat ticket creation; Power BI usage dashboard.
  • Validated with the L1 support team, resulting in fewer tickets raised.
Azure OpenAILangGraphFastAPIMarqoOpen WebUIMicrosoft Graph APIPower BI
Production

Water Pump Defect Prediction

Kalman filter-based predictive model on real-time time-series test data.

  • Earlier defect detection and reduced test time.
Kalman FilterTime-Series
Proof of Concept

Actuator Endurance Failure Prediction

LSTM model predicting the cycle count at which actuators fail, from endurance test logs.

LSTMTime-Series

Education

M.E., Systems Engineering and Operations Research

College of Engineering, Guindy (CEG), Anna University, Chennai

2018 · 79%

B.E., Computer Science and Engineering

Meenakshi Academy of Higher Education and Research, Chennai

2015 · 80%

Get in Touch

Let's Work Together

I'm always open to discussing new projects, creative ideas or opportunities to be part of your visions.

Location

Chennai, India | Open to Relocation

"Data is the new oil. It's valuable, but if unrefined it cannot really be used."

— Clive Humby