Phoenix Softnet Technologies - India

Senior Software Engineer

Ocean IT Solutions

Full Stack Developer

𝐀𝐈 𝐄𝐧𝐭𝐡𝐮𝐬𝐢𝐚𝐬𝐭 | 𝐅𝐮𝐥𝐥 𝐒𝐭𝐚𝐜𝐤 𝐃𝐞𝐯𝐞𝐥𝐨𝐩𝐞𝐫 | 𝐆𝐫𝐨𝐰𝐭𝐡 𝐌𝐚𝐫𝐤𝐞𝐭𝐞𝐫 | 𝐒𝐡𝐚𝐫𝐢𝐧𝐠 𝐈𝐧𝐬𝐢𝐠𝐡𝐭𝐬 𝐨𝐧 𝐀𝐈 | 𝐓𝐞𝐜𝐡𝐧𝐢𝐜𝐚𝐥 𝐂𝐨𝐧𝐭𝐞𝐧𝐭 𝐖𝐫𝐢𝐭𝐞𝐫

As a software developer with 3 years of experience, I am passionate about crafting high-quality software that solves complex problems and delivers value to end-users. With a strong background in Flutter,ASP .NET and Database technologies, I thrive on collaborating with cross-functional teams to design and implement solutions that meet business requirements.

My expertise spans the entire software development lifecycle, from requirements gathering and architecture design to coding, testing, deployment, and maintenance. I am comfortable working with a wide range of programming languages, frameworks, and tools, and am always eager to learn new technologies to expand my skill set.

If you are looking for a software developer who can help you achieve your business goals, please don't hesitate to get in touch. I am excited to explore new opportunities and to leverage my skills and experience to deliver value to your organization.

Aakash K.

Ford Credit

● Led data modernization project with GCP for Europe region at Ford Credit, ensuring alignment with regulatory restrictions.
● Developed data ingestion, analytical domain, and data mart for presentation layer, enhancing data accessibility and insights.
● Implemented solutions to optimize data processing and storage, improving efficiency and scalability.

Senior Data Engineer

Just Eat Takeaway.com

● Developed and implemented a Security Data Insights Project on GCP, facilitating real-time queries and resulting in a 30% reduction in response time for security threat mitigation.
● Collaborated cross-functionally with InfoSec teams to architect a scalable data infrastructure, leading to a 50% increase in data processing efficiency.
● Orchestrated the delivery of a dynamic dashboard and automated ETL pipeline, optimising vulnerability management processes and achieving a 40% improvement in data accuracy

Data Engineer

bigspark

● Led a team of 5 to create an MVP for a Business-Screening Project, resulting in a 20% increase in client satisfaction and a 30% reduction in time-to-market.
● Managed a team of 6 and collaborated with 20 pipeline owner teams to migrate 140 pipelines from Kafka Cloudera to Kafka Confluent, achieving a 50% reduction in data processing costs.
● Designed and constructed an end-to-end ingestion and analytics platform in Microsoft Azure for a Real Estate client, resulting in a 25% increase in data accessibility and a 15% improvement in data analysis speed.

PayU

● Automated and fine-tuned a critical machine learning model using Python, AWS EMR, Apache Spark, and Apache Airflow, resulting in a 30% improvement in model accuracy and a 20% reduction in model training time.
● Streamlined data processing by developing efficient data loading and aggregation frameworks using Apache Airflow, leading to a 40% decrease in data processing errors.
● Revitalised data and pipeline architecture, resulting in a 60% increase in business productivity and a 25% reduction in infrastructure costs.

● Collaborated with clients to identify process improvements for consumer payment options, resulting in a 15% increase in transaction success rates and a $10,000 reduction in payment processing costs.
● Reduced payment processing failure rates, saving $10,000 in costs.
● Devised and executed solutions that streamlined workflows and maximised resource utilisation, resulting in a 20% increase in team efficiency.

Senior Data Analyst

● Achieved up to a 23% improvement in the success rate of statistical analysis models on large data sets, leading to more accurate business insights.
● Revamped reporting dashboards and planning tools, reducing data processing time by 95% and enabling faster decision-making.
● Systematized data pipelines, reducing processing time by 40% and enhancing data accessibility for stakeholders.

Data Analyst

SAMRAT FORGINGS LIMITED

● Devised and deployed a modernised payroll management system utilising SQL and PHP, resulting in a 20% reduction in processing time for the HR team.

Intern

Security Data Engineer @Just Eat Takeaway| Ex-bigspark, PayU| Mentor @ Learnbay| Women Health | Mental Health

As the Senior Data Engineer at Ford Credit, my role encompasses leveraging advanced cloud technologies to drive innovation and efficiency in data management. Harnessing a solid foundation from Dr. A.P.J. Abdul Kalam Technical University in Computer Science, I've translated academic concepts into tangible results, such as implementing a Security Data Insights Project that significantly reduced threat response times. 

My commitment extends beyond technical skills to championing diversity, equity, and inclusion in the tech industry. As a mentor and advocate for Women in Tech, we work towards creating inclusive environments where every voice is heard and valued. This holistic approach to technology and society underscores my role as a leader and change-maker in the realm of data engineering and FinTech.

I actively engage in thought leadership and speaking opportunities, focusing on how inclusive practices can drive innovation in tech. Whether through keynote speeches, panel discussions, or collaborative projects, I aim to forge impactful partnerships that align with my commitment to innovation, DEI, and wellness.

I am eager to collaborate with forward-thinking organizations that share my vision for a more inclusive tech industry. Let’s connect to explore how we can work together to foster meaningful change and drive success through collaborative efforts.

Skills & Expertise:
Technical Skills: Data Engineering, Information Security, Cloud Data Architecture (Azure, AWS, GCP), Kafka, Apache Airflow, Python, SQL
Advocacy & Leadership: DEI Advocacy, Women’s Health, Mental Health, Public Speaking, Mentorship, Technical Project Leadership

Disha Mukherjee

Most engineers learn ACID in theory, 
But its importance becomes clear only when something breaks in production. 

Here is a simple reading-friendly explanation of each property.
1. Atomicity
A transaction is all or nothing. 
No partial writes. 
No half-completed updates. 
If one step fails, the entire transaction is rolled back and the system returns to a safe state.

2. Consistency
Every transaction must move the database from one valid state to another valid state. 
All rules, constraints and relationships must remain intact. 
This is different from the CAP theorem definition. 
Here, consistency is about preserving data integrity at every step.

3. Isolation
Multiple transactions often run at the same time. Isolation ensures they do not interfere with each other. 
The ideal model is serializability, where each transaction behaves as if it were running alone. 
In practice, databases use weaker isolation levels for performance, but the principle remains: data should not be confused about what happened first.

4. Durability
Once a transaction commits, it must remain stored. 
Even if the system crashes or a node fails. 
Modern systems achieve this through persistent logs, replication and stable storage.

Why ACID matters
ACID is not an academic concept. 
It is the reason money does not disappear mid-transfer. 
It is the reason orders are not duplicated. 
It is the reason data systems remain predictable and trustworthy at scale.

Image credit to ByteByteGo

Most data engineers learn tools first.
The ones who last in this field learn systems.

And the foundation of every reliable data system is a pipeline that follows five clear phases. 

After spending more than a decade building and fixing pipelines across different industries, I have found that teams struggle not because of technology, but because they don’t fully understand these phases.

Here is the structure behind every resilient data pipeline, regardless of scale or stack.

Collect
Everything begins with acquisition. Data arrives from devices, applications, logs, transactions, or third-party systems. If the collection process is inconsistent, you will spend the rest of the pipeline fighting avoidable issues.

Ingest
This is where raw data enters your ecosystem. It gets queued, validated, and organized. Ingestion failures ripple downstream, which is why strong ingestion practices are non-negotiable for operational stability.

Store
Once ingested, data needs to be stored in a way that supports both current and future use. Warehouses, lakes, lakehouses, databases. Storage is not about picking a tool. It is about designing for scale, access patterns, and cost.

Compute
 This is the heart of the pipeline. Aggregation, cleaning, transformations, partitioning, compression, batch jobs, stream processors. If the compute layer is slow or fragile, the entire system inherits that weakness.

Consume
The final phase is where the value shows up. Dashboards, analytics, operational use cases, ML features, decision engines. If consumption is painful, it is usually a sign that one of the earlier phases was not designed with intention.

When teams understand these phases clearly, they stop stitching tools together and start building pipelines with purpose. 

And that shift alone has a bigger impact on reliability than any new framework or platform.

Which of these five phases do you find your organization struggles with the most today?

Have you ever tried training a model locally and thought, 
“This will be quick”? 
And then ten minutes later you are staring at your screen thinking, 
“Why is the preprocessing step louder than the model?”

Anyone who has done real ML work knows the secret:
Local model training is not a neat, cinematic moment.
It is a full engineering workout.

Working with the Dell Pro Max Plus 16 has made that workout far more enjoyable, but the reality of the workflow is still honest.
Here is what local training actually looks like.

You discover bottlenecks you did not know you had
 → Your loader quietly becomes the villain
 → Your preprocessing decides to take a lunch break
 → Your feature pipeline shows you who is really in charge
The NVIDIA RTX PRO 5000 Blackwell GPU finally gives enough headroom for these truths to show up.

You adjust hyperparameters far more often than you admit
 → Batch size that “should work” does not
 → Learning rate needs a personality change
 → Model depth suddenly feels ambitious

The advantage of local GPU acceleration is simple.
You get to experiment before your coffee gets cold.

You start monitoring hardware like it is a stock market chart
VRAM usage spikes.
Data throughput dips.
CPU and GPU fight for attention.
Tools like nvidia-smi become a reflex.

This is where you realise model training is as much about engineering as it is about maths.

You meet every possible edge case
 → Strange samples
 → Schema surprises
 → Half-missing labels
 → Data that simply refuses to cooperate

 Running the entire pipeline locally reveals issues that distributed systems hide until it is too late.
You fail faster, which is a very good thing

That is the real power of local high-performance compute.
You learn quickly.
You fix quickly.
And your final model carries fewer surprises.

Local model training is not glamorous. 
It is messy, revealing, and incredibly educational.
And with the right hardware, it becomes something you can genuinely enjoy.

What about you?
What is the most unexpected challenge you have faced while training models locally?

Logan Lawler Cindy Olivo Crystal Cook

#DellProMax #DellTech #NVIDIA

After working across data engineering and ML systems for years, one thing has become very clear to me.
The right hardware does not just make things faster. It changes how you think, test, and build.

That is why the Dell Pro Max Plus 16 stands out for data engineers and AI or ML engineers who work end to end, not just at one layer of the stack.

Here is what makes the difference in real workflows.

Data engineering workloads rarely stay “just data engineering”
→ Ingestion turns into heavy preprocessing
→ Feature engineering grows more complex
→ Experiments start overlapping with model work
→ Pipelines need to behave closer to production
The Dell Pro Max handles this shift without forcing you to change how you work.

Local compute enables better pipeline design
→ Faster iteration on transformations
→ Honest debugging under realistic loads
→ Fewer shortcuts taken because something is “too slow to test locally”
This is where many pipeline issues surface early instead of in production.

The NVIDIA AI Platform matters more than people realise
It is not just about the GPU.
It is the combination of NVIDIA hardware and the AI software stack that enables efficient data processing, feature engineering, and model experimentation on the same system.
That full stack support is what makes local experimentation practical.

GPU acceleration is not only for training
→ Preprocessing large datasets
→ Feature generation at scale
→ Validation and profiling
→ Stress testing data logic
The NVIDIA RTX PRO 5000 Blackwell GPU gives enough headroom to treat these steps seriously instead of cutting corners.

One system that supports the entire workflow
→ Data ingestion and transformation
→ Feature engineering and validation
→ Model training and tuning
→ Debugging and experimentation
This reduces context switching between environments and keeps the feedback loop tight.

For data engineers moving closer to ML systems, and for ML engineers who care about data quality, this kind of setup matters.
It supports how modern systems are actually built, not how job descriptions describe them.

If you work across data pipelines and ML workflows, what part of your setup limits you the most today?

Logan Lawler Cindy Olivo Crystal Cook


#DellProMax #DellTech #NVIDIA

Today is International Mens Day.
A day that often gets far less conversation than it deserves.

We talk a lot about creating supportive workplaces, stronger communities, and healthier teams. But we rarely talk about the pressures many men carry silently while trying to meet those expectations.

In tech especially, I have seen incredibly talented men struggle quietly with:
 • performance pressure
 • fear of being judged for vulnerability
 • imposter syndrome in fast moving environments
 • the expectation to always appear strong or certain
 • work life imbalance they feel they "should not" talk about

The reality is simple.
Support is not gender specific.
Respect is not gender specific.
And growth definitely is not gender specific.

International Mens Day is a reminder to create spaces where everyone feels comfortable asking for help, admitting they are overwhelmed, or simply being human.

Celebrating the men who show up every day.
Who lead with integrity.
Who mentor, contribute, and hold things together even when it’s not easy.

Let’s build a culture where strength is not defined by silence, but by honesty and collaboration.

Everyone says ETL is old school. 
Outdated. 
Replaced by ELT.
But the reality inside real production systems tells a very different story.

Not every company has infinite compute.
Not every team has a cloud budget that can absorb poorly designed pipelines.
Not every workload is suited for dumping raw data into a warehouse and hoping transformations will magically scale.

ETL didn’t disappear. It just became unfashionable to talk about it.

In the real world, ETL is still the smarter choice when:
→ You need strict quality checks before data lands
→ You want to avoid polluting the warehouse with raw messy data
→ You have regulatory requirements that demand controlled transformations
→ Your data volume makes compute-heavy transformations expensive
→ You need predictable workloads instead of unpredictable warehouse bills
→ You’re ingesting data that isn’t analytics-ready without upstream cleaning

ELT is great for flexibility and exploration.
ETL is great for discipline and control.
Mature data teams know when to choose which.

The truth is simple:
The best architecture isn’t ETL or ELT.
It’s the one that respects your constraints, your cost, your latency expectations, and your team’s capability.

Tools evolve.
Hype cycles come and go.
But thoughtful engineering never goes out of style.


Follow Disha Mukherjee for Data Made Easy.

7 IN-DEMAND DATA ANALYTICS SKILLS 🚀 

Want to thrive in data analytics? Master these key skills to stay ahead in the industry! 

📌 Sql – extract, manipulate, and manage data efficiently 
📌 Spreadsheets – excel & google sheets for data handling and analysis 
📌 Statistical programming – python, r, and sas for advanced analytics 
📌 Data visualization – tableau & power bi to present insights effectively 
📌 Machine learning – algorithms like linear regression, decision trees & random forest 
📌 Database management – mysql, postgresql, sql server for handling large datasets 
📌 Soft skills – communication, critical thinking & creativity for impactful storytelling 

Upgrade your skills and unlock new opportunities in data analytics! 📊✨ 

♻️ 𝙍𝙚𝙥𝙤𝙨𝙩 𝙩𝙝𝙞𝙨 𝙞𝙛 𝙮𝙤𝙪 𝙛𝙞𝙣𝙙 𝙞𝙩 𝙝𝙚𝙡𝙥𝙛𝙪𝙡.

Let’s talk about your career goals. Book a call with me today! Link is in the Comment below!👇🏻👇🏻👇🏻.



#dataanalytics #sql #machinelearning #datavisualization #careergrowth #datawithdisha

Become a Top Data Engineer in 2025 & Dominate 2026!🚀

Want to build a high-paying career in Cloud Data Engineering? 

Here’s the roadmap suggested by Education Ellipse
Core Skills to Master
→ 𝐏𝐫𝐨𝐠𝐫𝐚𝐦𝐦𝐢𝐧𝐠 & 𝐐𝐮𝐞𝐫𝐲𝐢𝐧𝐠: SQL / Python
→ 𝐂𝐥𝐨𝐮𝐝 𝐅𝐮𝐧𝐝𝐚𝐦𝐞𝐧𝐭𝐚𝐥𝐬: Azure / AWS basics
→ 𝐂𝐥𝐨𝐮𝐝 𝐒𝐞𝐫𝐯𝐢𝐜𝐞𝐬: Azure / AWS Storage, Key Vault / KMS
→ 𝐄𝐓𝐋 & 𝐎𝐫𝐜𝐡𝐞𝐬𝐭𝐫𝐚𝐭𝐢𝐨𝐧: Azure Data Factory / AWS Glue
→ 𝐃𝐞𝐯𝐎𝐩𝐬: Azure DevOps / AWS CodePipeline
→ 𝐃𝐚𝐭𝐚 𝐖𝐚𝐫𝐞𝐡𝐨𝐮𝐬𝐢𝐧𝐠: Synapse / Redshift / Snowflake
→ 𝐁𝐢𝐠 𝐃𝐚𝐭𝐚 & 𝐀𝐧𝐚𝐥𝐲𝐭𝐢𝐜𝐬: Databricks / Spark / PySpark
→ 𝐃𝐚𝐭𝐚 𝐀𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐞: Delta Lake & Lakehouse
→ 𝐏𝐫𝐚𝐜𝐭𝐢𝐜𝐚𝐥 𝐒𝐤𝐢𝐥𝐥𝐬: End-to-End Project Implementation
→ 𝐂𝐚𝐫𝐞𝐞𝐫 𝐏𝐫𝐞𝐩: Resume Building & Interview Preparation

Now, if you want to become a top Cloud Data Engineer in 2025, Education Ellipse is conducting 𝐅𝐑𝐄𝐄 𝐌𝐚𝐬𝐭𝐞𝐫𝐜𝐥𝐚𝐬𝐬𝐞𝐬 on Azure, AWS and Snowflake Data Engineering 🌩️

Topic - How to become Highest Paid Cloud Data Engineer in 2026

𝐉𝐨𝐢𝐧 𝐨𝐮𝐫 𝐖𝐡𝐚𝐭𝐬𝐀𝐩𝐩 𝐆𝐫𝐨𝐮𝐩 𝐟𝐨𝐫 𝐌𝐚𝐬𝐭𝐞𝐫𝐜𝐥𝐚𝐬𝐬: 👉 https://lnkd.in/dUc4uHvP

𝐀𝐖𝐒 𝐃𝐚𝐭𝐚 𝐄𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠 𝐌𝐚𝐬𝐭𝐞𝐫𝐜𝐥𝐚𝐬𝐬
Date: 11th Dec 2025
Time: 7:00 PM IST
𝐑𝐞𝐠𝐢𝐬𝐭𝐞𝐫 𝐡𝐞𝐫𝐞 for Masterclass Link: https://lnkd.in/dYp4g9mS

𝐀𝐳𝐮𝐫𝐞 𝐚𝐧𝐝 𝐁𝐢𝐠 𝐃𝐚𝐭𝐚 𝐄𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠 𝐌𝐚𝐬𝐭𝐞𝐫𝐜𝐥𝐚𝐬𝐬
Date: 12th Dec 2025
Time: 7:00 PM IST
𝐑𝐞𝐠𝐢𝐬𝐭𝐞𝐫 𝐡𝐞𝐫𝐞 for Masterclass Link: https://lnkd.in/dbR6a8qw

💬 𝐒𝐭𝐚𝐲 𝐮𝐩𝐝𝐚𝐭𝐞𝐝 → Join our WhatsApp Channel:
 👉 https://lnkd.in/dVKamDuK

P.S. Share this with anyone who dreams of becoming a Cloud Data Engineering Rockstar in 2025! 💥

Ever opened a Data Engineer interview question sheet and thought-
 “Wait… they’re not just testing my tech skills, they’re testing how I think?”
You’re absolutely right. 

Here’s the truth:
Most data engineering interviews don’t just evaluate whether you can write SQL or build a data pipeline.

 They assess:
→ How you connect business problems to data systems
→ How you make design trade-offs
→ How you communicate your process

I’ve broken down the most common Data Engineer Interview Questions- not just with answers, but with the why behind them.

You’ll find:
 ↳ “General” questions that test your story and clarity
 ↳ “Process” questions that reveal your decision-making
 ↳ “Technical” questions that test depth, not just recall

Some highlights 
→ How to explain your “path to data engineering” without sounding like a
    CV
→ The right way to talk through a project from start to finish
→ What to say when they ask about schemas, unstructured data, and
   Hadoop (without jargon dumping)
→And the golden closer: how to ask them the right questions back.
Whether you’re just preparing for your first interview or aiming for a senior role, this guide will help you sound confident, structured, and human.

I’ve compiled the full breakdown (with examples, frameworks, and prep strategy) into a handy PDF:
 “The Ultimate Data Engineer Interview Prep Guide.”

#DataEngineering #CareerInData #TechInterviews #DataEngineer #SQL #WomenInTech #CareerGrowth

DE Interview Prep Guide.

The Dell Pro Max Plus 16 has completely changed how I approach the early stages of AI development.

Not during training.
Much earlier.
Inside the data pipeline where the real work begins.

Most conversations around AI still focus on the model.
But anyone who has built systems at scale knows this truth:

Your model is only as strong as the pipeline that feeds it.
And your pipeline is only as strong as your ability to 
→ Experiment, 
→Validate, and 
→ Refine without friction.

With the NVIDIA RTX PRO 5000 Blackwell GPU, I can treat preprocessing and feature engineering with the same seriousness as model training.

Here is the workflow I follow.
Start with the raw data, not assumptions
 → Pull the source exactly as it exists
 → Keep every inconsistency, missing value, and edge case
 → Let the dataset tell you how the system behaves
With the NVIDIA RTX PRO 5000 Blackwell GPU, repeated exploratory passes across large datasets become practical instead of painful.

Iterate on transformations in tight loops
 → Test encoding strategies
 → Compare normalization choices
 → Experiment with aggregations and text cleaning
 → Validate statistical transforms
GPU acceleration makes these operations fast enough that iteration becomes a natural part of thinking, not a blocker.

Validate outputs as deeply as the model
 → Look for distribution shifts
 → Test edge cases
 → Simulate heavier loads locally
 → Compare intermediate states
Most ML failures start here, not in the models. Local GPU compute lets me surface these issues early.

Package features only when the data is stable
 → Reproducible
 → Documented
 → Aligned with production behavior
 Once this foundation is stable, training becomes straightforward instead of reactive.

What this setup gives me is clarity.
Faster feedback loops.
Cleaner datasets.
And fewer surprises downstream.

If you work on feature engineering, I am curious:
Which step in your preprocessing pipeline takes the most time right now?

Logan Lawler Cindy Olivo Crystal Cook

#DellProMax #DellTech #NVIDIA

Every modern data system begins with one simple question:
How will your data get in?

That’s where data ingestion comes in; 
The backbone of every pipeline.

It’s not glamorous, but it’s where scalability, reliability, and cost are decided long before any dashboard comes to life.
Most teams think ingestion is “just about connecting sources.”

In reality, it’s about handling variety, velocity, and volume; all at once.

Some of the tools leading this space:
→ Azure Event Hubs for high-speed event collection and real-time analytics
→ Google Cloud Pub/Sub for asynchronous messaging and streaming ingestion
→ Apache Kafka for large-scale, low-latency event pipelines
→ Apache Flume for distributed log and event collection

Each serves the same purpose; moving data, but each does it differently, depending on your architecture and scale.

Here’s the real challenge:
Picking the right tool is not about brand preference.
It’s about understanding your data flow, latency needs, and consumption pattern.

Because when ingestion goes wrong, everything downstream suffers; from processing to analytics to trust.

If you’re building or optimizing data pipelines right now, which ingestion tool have you found most reliable in production?

Follow Disha Mukherjeefor Data Made Easy!

#DataEngineering #StreamingData #SystemDesign #Kafka #CloudComputing #ETL #DataPipeline

Image Credit Simform.

Good data products don’t start with SQL. 
They start with questions.

I’ve seen teams jump straight into models, schemas, and pipelines because the requirement looked “simple.”

 A new metric.
 A new dashboard.
 A new table.

But every time a data project struggles, the root cause is almost always the same:
No one slowed down to ask the right questions.

Before a single line of SQL is written, great data teams ask:
→What problem are we trying to solve?
 Not “what table should we build,” but “what decision will this enable?”

→ Who will use this?
 A data scientist, a product manager, a finance analyst — each needs a different kind of truth.

→ What does ‘correct’ look like?
 If the business doesn’t know the definition, the data team will guess, and that guess becomes debt.

→ What could go wrong later?
 A quiet discussion today prevents a loud incident tomorrow.

Once the questions are clear, SQL becomes the easy part.
The complexity drops.
The ownership becomes obvious.
The data product actually solves something.
Good data engineering is not about writing queries faster.
It’s about creating clarity before creating tables.

Takeaway:
 If you want better data products, start with alignment. The code follows naturally.

Follow Disha Mukherjee fore data made easy!

I am excited to announce that 
I have become a Dell Pro Max Ambassador.

As someone who has spent more than a decade building data systems and, more recently, working toward full-stack AI development, access to reliable high-performance compute has become a critical part of my workflow. 

Training models, running iterative experiments, and evaluating data-intensive pipelines require both speed and stability at the hardware and software level.

I have received the Dell Pro Max Plus 16, a system designed for professional-grade workloads. It is equipped with the NVIDIA RTX 5000 GPU built on the Blackwell architecture with 24 GB of GPU memory. 

This GPU is part of the NVIDIA AI Platform, which combines accelerated hardware with the NVIDIA AI software stack to support modern AI, simulation, and data workflows end to end.

Along with the system, 
I will be using the Dell UltraSharp Webcam, 
the Dell Pro Plus Compact Keyboard and Mouse (KM7120W), 
and Dell headphones to create detailed, workflow-driven content for the community.

Over the next few weeks, I will share practical, hands-on experiences around:
• Building and optimizing data pipelines
 • Training and fine-tuning models using local accelerated compute
 • Evaluating performance gains and iteration speed
 • Demonstrating how the NVIDIA AI Platform supports real AI development work
 • Offering guidance that other engineers and aspiring practitioners can apply directly

My goal is to produce content that is honest, useful, and grounded in real engineering practice, not theory.

Thank you to Dell Technologies for the opportunity to contribute to the program. 

More updates soon.

Logan Lawler Cindy Olivo Crystal Cook

 #DellProMax #DellTech #NVIDIA

Most people assume everything in AI has to run in the cloud.
But anyone who has worked with large datasets knows the cloud is only part of the workflow.
When we are preparing data for AI systems, the work is rarely linear.

You iterate.
You validate.
You experiment.

And every cycle depends on how quickly your environment responds.
Over the past decade in data engineering, I’ve seen teams move from on-prem clusters to cloud-native ecosystems. 

Both have their strengths, but neither eliminates the need for fast, predictable local compute. 
Especially when we are engineering features, exploring data at scale, or testing preprocessing strategies that directly influence model behaviour.

Cloud supports scale.
Local systems support understanding.

When we are preparing large datasets, a predictable environment becomes critical. 
You need to run transformations repeatedly and reliably. 
You need consistency when debugging subtle issues. 
And you need the freedom to iterate without waiting for resources or watching budgets fluctuate.

The Dell Pro Max Plus 16, equipped with the NVIDIA RTX PRO 5000 Blackwell GPU with 24GB of VRAM, brings that consistency back. 
It allows me to run preprocessing pipelines, feature engineering workloads, and training experiments without delays, queues, or variable performance.
For AI teams, iteration speed shapes outcomes.

The faster we can prepare, test, and refine data, the better our models become.
Quality in AI starts with the quality and reliability of the data workflow.
Local accelerated compute removes friction.

It keeps the entire process grounded and controlled.
And it allows deeper technical work to happen without interruption.

In the next posts, I will break down how this system supports each part of the data-to-model lifecycle, and where it makes the biggest difference.

What part of your workflow slows down your iteration speed the most?

Logan Lawler Cindy Olivo Crystal Cook

#DellProMax #DellTech #NVIDIA

AI won’t replace data engineers, but mediocre thinking will

Every few months someone confidently says that AI is about to automate the entire data engineering role.
But the reality is different.

AI can generate SQL.
AI can clean logs.
AI can suggest models.
AI can accelerate execution.

What AI cannot do is think.

It cannot understand why a business metric is broken.
It cannot decide the right trade off between cost and latency.
It cannot design a system that ages well under real world chaos.
It cannot ask the questions that shape the right solution.

The engineers who will thrive in an AI driven world are not the ones who know the most tools.
They are the ones who think clearly about problems.

They understand the domain.
They simplify the architecture.
They prioritise what matters.
They question assumptions.
They design for reliability, not just delivery.

AI amplifies this kind of thinking.
It makes good engineers faster.
But it also makes weak thinking visible very quickly.

If the foundation is unclear, AI only accelerates the confusion.

The future of data engineering is not about competing with AI.
It is about working alongside it with stronger judgement, sharper reasoning, and better questions.

AI will not replace data engineers.
But engineers who outsource their thinking will replace themselves.


Follow Disha Mukherjee for Data Made Easy.

Will AI Take Data Engineering Jobs

If you want to build a strong career in tech, especially Abroad, 
Focus on three things:
→ Personal branding.
→ Networking.
→ And asking the right questions.

Your personal brand helps people find you.
Your network helps people trust you.
And your questions help people remember you.

Too many skilled engineers stay invisible because they never show what they know.
Too many ambitious professionals connect without context, adding people, not building relationships.

Start small.
Share your perspective on the problems you solve.
Engage with others’ posts thoughtfully.

Reach out with intent- not to ask for a job, but to learn.
One thoughtful conversation can open more doors than ten cold applications.

And one consistent voice online can build more credibility than years of quiet expertise.
Your skills are your foundation.

Your brand and network are your amplifier.

#CareerInTech #DataEngineering #PersonalBrand #Networking #TechLeadership #UKTech #DataEvangelism

The cost of not deleting anything: data debt

Most data teams think their problems come from complex pipelines or slow queries.
But a huge part of the struggle comes from something much simpler.
Not deleting things.

Old tables that no one owns.
Metrics created for projects that ended long back.
Pipelines that still run even though no one knows why.
Dashboards that have not been opened in months.

All of this becomes data debt.
And data debt grows quietly.

You feel it during migrations.
Every change takes longer because you need to check what might break.
You feel it during outages.
Debugging becomes harder because no one knows which assets matter.
You feel it during onboarding.
New engineers spend more time decoding history than creating impact.

Most data debt starts with good intentions.
A quick table.
A temporary metric.
A one time pipeline.
The team moves fast.
But nothing gets retired.

The fix is not a one time cleanup.
The fix is a habit.
Review. Remove. Archive. Repeat.

If your data stack feels slow, cluttered, or risky, you may not need more tools.
You may need fewer unused assets.

Delete what no longer serves the work.
Your future team will thank you.

Most people prepare for Data Engineering interviews 
By memorizing answers.
But interviews aren’t about reciting definitions.

They’re about showing how you think, how you solve problems, and how well you understand the flow of data.

If you’re preparing for DE roles in 2025, 
There are a few patterns you’ll notice.

Companies want engineers who can design systems, 
Reason about trade-offs, and communicate clearly.

Tools matter, 
But clarity matters more.

I spent some time reviewing the most common interview questions asked across companies.

The themes are consistent: data modeling, pipeline design, orchestration, Python, SQL, scaling decisions, and how you collaborate across teams.

 And almost every interview includes one deeper question:
 “Walk me through a data project you designed end to end.”
Most candidates fail here because they talk about tools instead of the problem they solved.

If you want to stand out, focus on three things:
→Show how you think.
→Break down a problem. 
→Explain the constraints. 
→Share why you chose one approach over another.
→Show how data moves.

Interviewers want to see how you design reliable pipelines, ensure data quality, and handle failure scenarios.

Show your communication muscles.
A lot of data engineering is working with analysts, PMs, ML teams and explaining trade-offs in plain English.

To make preparation easier, Use interview prep guide covering the most asked DE questions:
HR screen, technical deep dives, SQL, Python, orchestration, big data, and system design.

What interview topic do you find the most challenging right now? 
Let’s discuss in the comments.

Follow Disha Mukherjee for Data Made Easy!

Data Engineering Interview Prep.

Most people think GPUs start mattering only when you begin training models.
The reality is very different.

The ceiling of any AI system is set long before training.
It is set in the pipeline.
If your data is slow, inconsistent, or poorly transformed, no model can recover from that.

After more than a decade in data engineering, this is the truth I keep coming back to:
Pipeline performance shapes model performance long before the first epoch runs.

Working with the 𝐃𝐞𝐥𝐥 𝐏𝐫𝐨 𝐌𝐚𝐱 𝐏𝐥𝐮𝐬 16 has reminded me how much local high-performance compute changes that early stage. It is not just about faster results. It is about faster iteration. And iteration is where real engineering progress happens.

Here are the areas where local GPU acceleration has made a very real difference in my workflow.

𝐃𝐚𝐭𝐚 𝐢𝐧𝐠𝐞𝐬𝐭𝐢𝐨𝐧 𝐚𝐧𝐝 𝐩𝐫𝐞𝐩𝐫𝐨𝐜𝐞𝐬𝐬𝐢𝐧𝐠
Large unstructured datasets usually bottleneck at parsing and cleaning.
Offloading parts of this work to the 𝐍𝐕𝐈𝐃𝐈𝐀 𝐑𝐓𝐗 𝐏𝐑𝐎 5000 𝐁𝐥𝐚𝐜𝐤𝐰𝐞𝐥𝐥 𝐆𝐏𝐔 gives me quicker feedback on assumptions, schema issues, and anomalies.
Feature engineering at scale

𝐅𝐞𝐚𝐭𝐮𝐫𝐞 𝐠𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐨𝐧 𝐢𝐬 𝐞𝐱𝐩𝐞𝐧𝐬𝐢𝐯𝐞.
With GPU-accelerated libraries, I can try transformations that would normally be too slow on CPU-only setups. This changes the quality of what you even attempt.

𝐏𝐢𝐩𝐞𝐥𝐢𝐧𝐞 𝐢𝐭𝐞𝐫𝐚𝐭𝐢𝐨𝐧 𝐚𝐧𝐝 𝐝𝐞𝐛𝐮𝐠𝐠𝐢𝐧𝐠
When you can simulate production-like loads locally, you debug differently.
You catch edge cases earlier.
You stress test logic honestly.
And you do all of it without burning cloud budget.

Here is why this matters.
A clean, well-structured pipeline is not optional.
It is the foundation of every reliable AI system.
Improving it locally reduces training time, lowers cost, and prevents redesign later.

For those working with large-scale data workflows, I’m curious:
How do you handle pipeline experimentation and debugging today?

Logan Lawler Cindy Olivo Crystal Cook

#DellProMax #DellTech  #NVIDIA

When I first started using dbt Labs, it felt like just another transformation tool.
Write SQL, run models, build tables- simple enough.

But over time, I realized something deeper.
dbt wasn’t just about transformations.
It was teaching data teams how to think like product engineers.

Here’s how:
1. 𝐕𝐞𝐫𝐬𝐢𝐨𝐧 𝐂𝐨𝐧𝐭𝐫𝐨𝐥 𝐚𝐬 𝐚 𝐌𝐢𝐧𝐝𝐬𝐞𝐭
 Every model lives in Git. Every change is reviewed.
 That single shift turns data from a “shared folder of tables” into a product
 with lifecycle management.

2. 𝐓𝐞𝐬𝐭𝐢𝐧𝐠 𝐚𝐬 𝐚 𝐅𝐞𝐚𝐭𝐮𝐫𝐞, 𝐍𝐨𝐭 𝐚𝐧 𝐀𝐟𝐭𝐞𝐫𝐭𝐡𝐨𝐮𝐠𝐡𝐭
 In dbt, tests aren’t optional.
 They’re part of the build.
 Just like a product, your data ships only if it passes quality gates.

3. 𝐃𝐨𝐜𝐮𝐦𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧 𝐚𝐬 𝐚 𝐅𝐢𝐫𝐬𝐭-𝐂𝐥𝐚𝐬𝐬 𝐂𝐢𝐭𝐢𝐳𝐞𝐧
 Auto-generated docs may sound simple, but they redefine ownership.
 When every model is documented in context, data stops being tribal 
 knowledge.

4. 𝐓𝐫𝐚𝐧𝐬𝐩𝐚𝐫𝐞𝐧𝐜𝐲 𝐚𝐬 𝐭𝐡𝐞 𝐃𝐞𝐟𝐚𝐮𝐥𝐭
 With lineage, every dependency is visible.
 Suddenly, consumers can see how data is built, not just trust that it works.

This is what “data as a product” looks like in action:
↳Ownership
↳Quality
↳Discoverability
↳Feedback

Tools like dbt Labs don’t just change how we engineer data.
They change how we think about data.
And that mindset shift is what modern data teams are built on.

Takeaway:
The best data tools don’t just solve technical problems, they shape better engineering behavior.

#DataEngineering #dbt #DataAsAProduct #ModernDataStack #DataCulture #DataEvangelism

taplio

The Battle

Aakash K.

Disha Mukherjee

Followers

Posts

Engagements

Likes

Performance ScoreDisha Mukherjee's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Aakash K.

Disha Mukherjee

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Disha Mukherjee's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreDisha Mukherjee's LinkedIn Performance