Course

You will find below course syllabus and schedule for AI-839: ML Production Engineering course offered at IIIT-B in the Fall of 2024. This will evolve as the course progress.

Syllabus & Schedule

Part-1: Essentials
Wk Dt Topics Resources
01 01-Aug Introduction
1. Course Intro
2. ML in Production - Intro
W01-L01
  HW-01 Due Tue, 13th, August, 2024, 11.59pm IST
02 06-Aug
09-Aug		 Discovery
1. Contd. ML in Production - Intro
2. Software 1.0 vs 2.0
3. Design Thinking, Project Cards
4. MLOps Canvas & Tool Landscape		 W02-L01
W02-L02
  HW-02 Due Fri, 23rd, August, 2024, 11.59pm IST
03 13-Aug
16-Aug		 Models for Modeling
1. Kedro
2. DevOps		 W03-L01
W03-L02
  Note: 16-Aug DevOps guest lecture by Ravi
04 20-Aug
23-Aug		 Dev Setup & Data Monitoring
1. Kedro, PyTest, Ruff, Quartodoc
2. Data Quality		 W04-L01
W04-L02
  HW-03
HW-04		 11.59PM IST, Friday, 6th Sep, 2024.
11.59PM IST, Friday, 16th Sep, 2024.
05 27-Aug
30-Aug		 Model Monitoring and Deployment
1. Monitoring Metrics, Tests, Label Generation
2. Deployment with MLFlow		 W05-L01
W05-L02
 
06 03-Sep
06-Sep		 Evaluation and Governance
1. Hypothesis Tets, DoEs, Model Comparison
2. The R4 Framework		 W06-L01
W06-L02
  HW-05
HW-06		 Due 11.59PM IST, Tuesday, 24th Sep, 2024
Due 11.59PM IST, Friday, 18th Oct, 2024
Part-2: Full Stack ML
Wk Dt Topics Resources
07 10-Sep
13-Sep		 Scaling Laws, Sample Hardness
1. Sample Sizes, Active Learning, Scaling Laws
2. Sample Hardness		 W07-L01
W07-L02
  Minor Project Due 11.59PM IST, Friday, 25th Oct, 2024
08 17-Sep
20-Sep		 Sample Fitness, Guest Lecture
1. Likelihood Ratio, \(\nu\text{-information}\)
2. ML Platform		 W08-L01
Talk by Abhishek
 
09 24-Sep
27-Sep		 UQ, Midterm
1. Uncertainty Quantification, Conformal prediction
2. Midterm		 W09-L01
NA
  Midterm in-class
10 15-Oct
18-Oct		 Edge Deployment, ML Platforms
1. Model Compression, Quantization
2. Distributed Systems, Solving ML Eng. issues with Keras and TensorFlow		 Talk by Dr. Srinivas
Talk by Kalyan
 
11 22-Oct
25-Oct		 Causal ML, Robustness
1. DoWhy and DICE
2. Gradients is all you need		 Talk by Dr. Amit
W11-L02
 
12 29-Oct
01-Nov		 Robustness
1. Securing ML
2. Holiday		 Talk by Manoj Parmar
Text Attacks
Adversarial Attacks
 
13 07-Nov
08-Nov		 LLMs
1. LLMs Intro
2. LLMOps
3. Fullstack LLMs		 W13-L01
W13-L02
W13-L03
 
14 12-Nov
15-Nov		 Talks
1. Lessons Learnt in Production
2. Building datasets		 Talk by Dr. Venkata
Talk by Puneet
 
15 19-Nov
22-Nov		 Meta Learning
No Class		 W15-L01
 
16 Last Week Student Presentations
  26-Nov

29-Nov
30-Nov		 Fairness & Bias
Machine Unlearning
In-class final presentations-1
In-class final presentations-2		 W16-L01
W16-L02
Guest Lectures
Wk Dt Topics Resources
03 16-Aug Intro to DevOps
[ML Engineering]		 Ravi Mula
Architect: Sanketika
08 20-Sep ML Platforms
[ML Engineering]		 Abhishek Choudhary
Co-Founder & CTO: TrueFoundry
10 15-Oct Deploying on Edges
[ML Engineering]
Deck		 Dr. Srinivas Rana
Sr ML Scientist: Wadhwani AI
10 18-Oct MLOps @ Scale
[ML Engineering]
Deck		 Kalyan Deepak
Sr Staff Engineer, LinkedIn
11 22-Oct DoWhy, DiCE
[Causal ML]
Deck		 Dr. Amit Sharma
Principal Researcher, MSR
12 29-Oct Securing AI
[Adversarial ML]		 Manojkumar Parmar
CEO & CTO AI Shield Bosch AI Shield
14 12-Nov Building LLMs in Production in Regulated Industry
[System Design]		 Dr. Venkata Pingali
Co-Founder: Scribble Data
14 15-Nov Problems and Solutions in Data Labelling
[Data Collection]		 Puneet Jindal
CEO, Labellerr

Discussions

We will use WhatsApp group for (informal)discussions and alerts.

References

  1. [book] ML Engineering, Andiry Burkov, 2019, LeanPub
  2. [book] Effective Data Science Infrastructure, Vile Tuulos, 2023, Manning
  3. [book] ML System Design, Chip Huyen, 2023, O’Reilly
  4. [course] CS329S @ Stanford: ML Systems Design, Chip Huyen, 2022
  5. [course] MLOps, Chip Huyen, 2024

Grading

  • 40%: Six assignments
  • 15%: Midterm mini project
  • 20%: In-class midterm
  • 25%: Capstone project

Policies

  • Late Submissions:  All deadlines are due at on the date and time indicated on the course page. The penalties for late submission are as follows:
    • Late submissions not allowed (incur a zero) - except with prior approval or in valid exceptional cases
  • Make-up Exam/Submission Policy: As per institute policy
  • Citation Policy for Papers: Always mention the source, give full attribution and credits to citations, and as per institute policy
  • Academic Dishonesty/Plagiarism: As per institute policy
  • Accommodation of Divyangs: As per institute policy

Soma S Dhavala
Course Instructor