Ganesh Jumbad | AI & Data

AI | Data Platform Engineer | MLOps + System Design
Engineering for Engineers 🧑‍💻

[Ganesh Jumbad | AI & Data @zummbad]

@claudeai OMG.. @AnthropicAI is bringing whole new office suite for users

[Ganesh Jumbad | AI & Data @zummbad]

@hamids I think the primary reason is that one should be trustworthy. Trust is the main factor this is happening now. If you can break the trust if your loved ones then no one else is above them and also this is being viewed as national security

[Ganesh Jumbad | AI & Data @zummbad]

We may never see NVIDIA and Jensen Huang together in the same way again.

A partnership that created history, transformed the tech industry, and reshaped the future of AI might one day become just a chapter in history.

What a run!

[NVIDIA Newsroom @nvidianewsroom]

NVIDIA announces CEO transition. Read the release: https://t.co/MTTUvFHCJA

[Ganesh Jumbad | AI & Data @zummbad]

@imlikhita @grok You’re cooked 😂

[Ganesh Jumbad | AI & Data @zummbad]

@WatcherGuru We may never see NVIDIA and Jensen Huang together in the same way again.

A partnership that created history, transformed the tech industry, and reshaped the future of AI might one day become just a chapter in history.

What a run!

[Ganesh Jumbad | AI & Data @zummbad]

@nvidianewsroom This was inevitable..! Once reputation is damaged, you cannot face the audience..!!

[Ganesh Jumbad | AI & Data @zummbad]

Follow @zummbad for more Insights on AI Assisted Data Engineering..!

[Ganesh Jumbad | AI & Data @zummbad]

6/8

But don't make the mistake of abandoning fundamentals.

AI agents still need:

→ Reliable data pipelines
→ Correct data models
→ Data quality and validation
→ Governance and access controls
→ Lineage and observability
→ Scalable distributed systems

An AI agent connected to poor-quality data doesn't become intelligent.

It becomes confidently wrong.

[Ganesh Jumbad | AI & Data @zummbad]

8/8

My advice to aspiring Data Engineers:

Don't chase every new AI framework.

First, become excellent at building reliable data systems.

Then learn how LLMs, agents and AI applications consume, retrieve and reason over that data.

The future isn't Data Engineering versus AI Engineering.

It's Data Engineering becoming one of the foundations of AI Engineering.

[Ganesh Jumbad | AI & Data @zummbad]

5/8

What does this mean for aspiring Data Engineers?

Your learning roadmap needs to expand.

Beyond SQL, Python, Spark, Kafka and Airflow, start understanding:

→ Vector databases and embeddings
→ RAG and retrieval systems
→ Semantic layers and knowledge graphs
→ LLM evaluation
→ Agentic workflows and MCP
→ AI-assisted engineering

[Ganesh Jumbad | AI & Data @zummbad]

1/8

Data Engineering is changing faster than many aspiring engineers realise.

I looked at recent job descriptions from Amazon, McDonald's and Thomson Reuters.

One thing is becoming clear:

Companies increasingly want Data Engineers who can build infrastructure for AI, not just pipelines for analytics.

Here's what is changing. 🧵

[Ganesh Jumbad | AI & Data @zummbad]

4/8

Thomson Reuters is hiring a Lead Data & AI Engineer in Bengaluru.

The role combines:

→ Data Engineering
→ Semantic and dimensional modelling
→ AI engineering
→ Agents and applications
→ Data products

The expectation is to own the journey from raw data to AI-enabled decisions.

Job: https://t.co/lJO4k2uXrg

[Ganesh Jumbad | AI & Data @zummbad]

3/8

Amazon's Data Engineer II role in Bengaluru goes further.

It explicitly mentions:

→ AI-native data infrastructure
→ Semantic layers and knowledge graphs
→ Multi-agent orchestration
→ GenAI-powered data quality
→ Real-time pipelines for AI/ML inference and training

This is a significant expansion of traditional Data Engineering.

Job: https://t.co/Om850dJAR0

[Ganesh Jumbad | AI & Data @zummbad]

2/8

Look at McDonald's hiring for a Data Engineer III in Hyderabad.

The role involves:

1. Kafka and real-time streaming
2. Feature-ready datasets for computer vision and AI agents
3. Data contracts and schema management
4. AI coding assistants for SQL, Python, dbt and testing

AI is becoming part of the everyday Data Engineering workflow.

Job: https://t.co/Q5veDPQtgy

[Ganesh Jumbad | AI & Data @zummbad]

Databricks is bridging all the gaps for Data and AI Engineering..!

[Databricks @databricks]

Introducing Databricks Genie plugins for ChatGPT, Claude, Microsoft Copilot Cowork and Grok Bot!

More data doesn't automatically mean better answers. An AI agent may find the right data but still use the wrong context.

Now, your agents can tap into Genie to access governed business context, investigate questions across data and documents, and return answers with citations, all within their existing workflow.

Build your Ontology once. Bring it to the AI tools your teams already use.

https://t.co/1ftNAVklZV

[Ganesh Jumbad | AI & Data @zummbad]

"Write a data quality checker" is not a complete engineering requirement.

A better specification defines:

Input: a dataset and a set of rules.

Checks:
→ Required columns exist
→ Primary keys are unique
→ Required values aren't null
→ Numeric values satisfy valid ranges
Output:
→ Failed checks
→ Affected row counts
→ Error details
→ A clear pass/fail result
Also define what happens when a check fails.
The more precisely you define expected behavior, the easier it becomes to test whether AI-generated code is actually useful.

[Ganesh Jumbad | AI & Data @zummbad]

@eurofounder Don’t tell me it’s real..!

[Ganesh Jumbad | AI & Data @zummbad]

Simple analogy:

A judge in a courtroom makes decisions based on the evidence presented by lawyers and the applicable laws.

If lawyers provide incomplete or misleading evidence, the judge may reach an incorrect verdict. But even with the right evidence, a judge can still make mistakes.

Similarly, an LLM-as-a-Judge needs the right context, clear evaluation criteria, and verifiable evidence to assess another LLM’s output.

But here’s the catch: The judge itself also needs to be evaluated.

That’s why LLM-as-a-Judge systems need calibration, consistency checks, and human oversight rather than blind trust.

[Ganesh Jumbad | AI & Data @zummbad]

AI can help document a Data Engineering project.

But there is a difference between documentation that sounds convincing and documentation that reflects reality.

Ask it to explain:
→ Input and output schemas
→ Transformation rules
→ Configuration options
→ Dependencies
→ Failure modes
→ Operational assumptions
Then compare the documentation with the implementation and tests.
A particularly useful check:
Can a new engineer understand how to run the project and troubleshoot a failure without asking you?
Good documentation should reduce that dependency.