Day 1 - Understanding MLOps v/s DevOps
👋 Hello! I'm passionate about DevOps and have over 1+ years of experience in the field. I'm proficient in a variety of cutting-edge technologies and always motivated to expand my knowledge and skills. Let's connect and grow together!
SKILLS:
🔹 Languages & Runtimes: Python, Shell Scripting, HCL, YAML 🔹 Cloud Technologies: AWS, Microsoft Azure, GCP 🔹 Infrastructure Tools: Docker, Terraform, AWS CloudFormation 🔹 Other Tools: Linux, Git and GitHub Actions, Jenkins, Jira, GitLab (beginner), Docker, AWS DevOps 🔹 Web Development: HTML, CSS, Bootstrap, Python, SQL
Job & Responsibilities:
🚀 Improved development efficiency by implementing CI/CD pipelines, resulting in a 30% reduction in deployment time on the test server. 🔒 Strengthened deployment and testing reliability by utilizing Docker containers and optimizing Dockerfile, reducing development issues on the test server by 20%. ⚙️ Automated S3 bucket log creation with Shell scripting, eliminating 100% of manual search and saving 2 hours per week. 📅 Scheduled EC2 instance start/stop using Lambda functions and Event Bridge, leading to a 25% decrease in infrastructure costs. 🔧 Utilized AWS, Linux, Python, Docker, Shell scripting, Terraform, Jenkins Pipelines, and automation to streamline workflows and improve overall system performance.
I'm very detail-oriented and possess strong written and verbal communication skills. As a high performer with a possibility mindset, I strive to solve problems using efficient approaches.
Let's Connect & Grow:
If you find my profile suitable for the role you are searching for, please feel free to reach out to me at sumanprasad9766@gmail.com.
1. What Is MLOps?
Think of a food delivery app like Swiggy or Zomato.
They use machine learning (ML) to:
predict delivery time
recommend restaurants
detect fraud
optimize delivery routes
But building an ML model once is not enough.
The real challenge is:
running it daily
updating it when data changes
checking it is correct
scaling it to millions of users
fixing it when it breaks
This is where MLOps comes in.
Simple definition
MLOps = DevOps + Machine Learning
Or even simpler:
MLOps is the process of building, deploying, monitoring, and improving ML models in real-world companies.
2. Why MLOps Is Needed (Real-World Problems)
Example 1 — Food Delivery ETA Prediction
If the model says every order will arrive in 10 minutes:
customers get angry
restaurants get complaints
drivers rush and cause risk
Without MLOps:
No one monitors if predictions go wrong
No way to retrain the model when traffic patterns change
Engineers manually deploy updates → slow and error-prone
With MLOps:
model is tested before release
monitored after release
retrained automatically
rolled back if broken
Example 2 — Bank Loan Approval
Banks use ML to approve or reject loans.
Without MLOps:
Model trained on old data → unfair results
No tracking → cannot explain why loan rejected
No auditing → legal risk
With MLOps:
every version tracked
data stored securely
bias checked
approvals auditable
Example 3 — Hospital Patient Risk Prediction
Hospitals use ML to predict ICU admission risk.
Without MLOps:
model drift = accuracy drops
lives may be at risk
With MLOps:
real-time monitoring
alerts if accuracy drops
auto retrain
3. MLOps vs DevOps (Simple Comparison
)
| Area | DevOps | MLOps |
| Main Goal | Deploy software reliably | Deploy ML models reliably |
| Key Output | Application code | Model + data + code |
| Versioning | Code | Code + Data + Model |
| Testing | Functional tests | Data quality + ML accuracy tests |
| Monitoring | Uptime, errors | Accuracy, drift, bias + uptime |
| Retraining | Not needed | Required regularly |
| Fail Risk | App crash | Wrong predictions |
So MLOps is DevOps plus many ML-specific responsibilities.
4. Core Components of MLOps (Plain English)
Think of an ML lifecycle like running a restaurant kitchen.
| Kitchen Step | MLOps Step |
| Buy ingredients | Collect data |
| Clean ingredients | Clean data |
| Decide recipe | Select model |
| Cook | Train |
| Taste test | Validate |
| Serve | Deploy |
| Collect feedback | Monitor |
| Improve recipe | Retrain |
5. References:
Good for beginners:
Made With ML — What is MLOps?
Weights & Biases — Intro to MLOps
Neptune.ai — What is MLOps?
