Lilly Tech Systems
Preparation Checklist & Timeline
A structured preparation plan is the difference between a scattered, stressful interview process and a confident, systematic one. Choose the timeline that matches your experience level and available time.
4-Week Intensive Plan
Best for: Experienced ML engineers with recent interview experience who need a focused refresher. Requires 3–4 hours per day.
| Week | Focus Area | Daily Activities | Goal |
|---|---|---|---|
| Week 1 | ML Theory & Fundamentals | Review all 30 theory questions from Lesson 4. Explain each concept out loud. Solve 2 ML coding problems daily. | Can explain any ML concept in under 2 minutes without notes |
| Week 2 | ML Coding & Implementation | Implement gradient descent, decision tree, neural network, k-means, and data pipeline from scratch. Do 2 LeetCode mediums daily (ML-adjacent: arrays, hashmaps, trees). | Can implement any standard ML algorithm in 30 minutes |
| Week 3 | System Design & Behavioral | Practice 1 full system design walkthrough daily (recommendation, fraud, search, ads, content moderation). Prepare 8 STARL stories. Do 1 mock behavioral interview. | Can fill 45 minutes on any ML system design topic with depth |
| Week 4 | Mock Interviews & Weak Areas | 2 full mock interviews (1 technical, 1 behavioral). Review and strengthen weak areas. Rest the day before each interview. | Simulate real interview conditions with timing and pressure |
8-Week Standard Plan
Best for: ML engineers who have not interviewed in 1–2 years. Requires 2–3 hours per day.
| Weeks | Focus Area | Key Activities |
|---|---|---|
| 1–2 | ML Foundations Rebuild | Review statistics, probability, linear algebra fundamentals. Study supervised/unsupervised/reinforcement learning distinctions. Reread key papers (Attention Is All You Need, ResNet, XGBoost). |
| 3–4 | ML Coding Fluency | Implement 5 core algorithms from scratch (gradient descent, decision tree, k-means, neural network, data pipeline). Solve 3 ML coding problems per day. Practice with timer (45 min per problem). |
| 5–6 | ML System Design | Study the 4-step framework. Practice 6 different ML system design problems. Study real-world architectures (Netflix recommendations, Uber pricing, Google search). Read ML engineering blogs. |
| 7 | Behavioral Preparation | Write 8 STARL stories covering all key themes. Practice telling each story in 2 minutes. Research target companies' values and leadership principles. Do 2 mock behavioral interviews. |
| 8 | Integration & Mock Interviews | 3–4 full mock interviews covering all round types. Review and patch weak areas. Prepare questions for interviewers. Rest and build confidence. |
12-Week Comprehensive Plan
Best for: Career switchers or engineers new to ML roles. Requires 3–4 hours per day.
| Weeks | Focus Area | Key Activities |
|---|---|---|
| 1–3 | ML Foundations | Complete an ML course (Andrew Ng's ML Specialization or fast.ai). Understand supervised, unsupervised, and deep learning fundamentals. Implement basic algorithms in Python. |
| 4–5 | Build ML Projects | Build 2–3 end-to-end ML projects to discuss in interviews. Include: data cleaning, feature engineering, model training, evaluation, and deployment. Document results and decisions. |
| 6–7 | ML Theory Mastery | Study all 30 theory questions deeply. Understand the math behind each concept. Practice explaining concepts to a non-technical friend. Create flashcards for quick review. |
| 8–9 | ML Coding Practice | Daily coding practice (2–3 problems). Focus on ML implementations and data manipulation. Build speed and confidence with timed sessions. |
| 10–11 | System Design & Behavioral | Study ML system design framework. Practice 8+ design problems. Prepare behavioral stories. Start mock interviews (2 per week). |
| 12 | Final Preparation | Full mock interview simulations. Review all weak areas. Prepare logistics (outfit, setup, questions). Mental preparation and confidence building. |
Curated Resource List
ML Theory & Fundamentals
- An Introduction to Statistical Learning (ISLR) — Free textbook. The gold standard for ML theory with intuitive explanations.
- Andrew Ng's Machine Learning Specialization (Coursera) — Best structured introduction to ML concepts.
- Stanford CS229 Lecture Notes — Free. More mathematical depth than Ng's course.
- The Hundred-Page Machine Learning Book by Andriy Burkov — Concise reference for interview review.
ML Coding Practice
- LeetCode (ML-tagged problems) — Filter by "Machine Learning" and "Data Science" tags.
- Kaggle — Practice data manipulation and feature engineering on real datasets.
- Implement from scratch: Linear regression, logistic regression, decision tree, random forest, k-means, KNN, neural network, gradient descent variants.
ML System Design
- Designing Machine Learning Systems by Chip Huyen — The best book on ML system design for interviews.
- Machine Learning Design Patterns by Lakshmanan, Robinson & Munn — Google's approach to ML architecture.
- Engineering blogs: Netflix Tech Blog, Uber Engineering, Airbnb Engineering, Google AI Blog — Real-world ML system designs.
Behavioral Preparation
- Amazon Leadership Principles — Study all 16 even if not interviewing at Amazon. Many companies use similar frameworks.
- Glassdoor interview reviews — Read recent interview experiences for your target company and role.
- Practice with a friend or mentor — Behavioral answers improve dramatically with rehearsal.
Mock Interview Tips
How to Run an Effective Mock Interview
- Find a partner. Another ML engineer preparing for interviews is ideal. You take turns being interviewer and candidate.
- Use real questions. Pull questions from this course, Glassdoor, or Blind. Do not use the same question twice.
- Enforce time limits. 45 minutes for coding, 45 minutes for system design, 30 minutes for behavioral. No extensions.
- Give structured feedback. After each mock, discuss: (a) What went well? (b) What needs improvement? (c) Would you have hired this candidate?
- Record yourself. Watch the recording to catch filler words, long pauses, and unclear explanations. This is uncomfortable but incredibly effective.
Self-Assessment Checklist
Before your interview, honestly rate yourself on each area (1–5 scale). Focus remaining prep time on anything below a 4:
| Skill | Can you do this? | Rating |
|---|---|---|
| Explain bias-variance tradeoff with an example | Without notes, in under 2 min | __/5 |
| Implement gradient descent from scratch | In 15 minutes, bug-free | __/5 |
| Design a recommendation system end-to-end | Fill 45 minutes with depth | __/5 |
| Explain precision vs recall and when to use each | With real-world examples | __/5 |
| Tell a STARL story about a failed ML project | In 2 minutes, with a clear lesson | __/5 |
| Handle "why this company?" question | With specific, researched reasons | __/5 |
| Discuss L1 vs L2 regularization | Including when to use each | __/5 |
| Write clean, documented Python code under pressure | With docstrings and edge cases | __/5 |
| Evaluate a compensation offer | Calculate 4-year total comp | __/5 |
| Explain attention mechanism in Transformers | Q, K, V, softmax, multi-head | __/5 |
Frequently Asked Questions
How many hours should I study per day?
2–3 focused hours per day is optimal for most people. Quality matters more than quantity. Use active practice (solving problems, doing mock interviews) rather than passive study (watching videos, reading textbooks). If you can only do 1 hour per day, extend your timeline accordingly.
Should I apply to my top-choice company first or last?
Apply to your top choice in the middle of your interview cycle. Use your first 1–2 interviews at lower-priority companies as warm-up. By the time you reach your top choice, you will have real interview experience and may already have competing offers to strengthen your negotiation position.
Do I need a PhD for ML engineer roles?
No. A PhD is valued for research scientist roles but is not required for ML engineer positions at most companies, including FAANG. Strong practical skills (deploying models to production, building data pipelines, writing clean code) are more valued than academic credentials. That said, a PhD candidate with publications may receive a higher initial level offer (L4 vs L3).
How important is LeetCode for ML interviews?
Less important than for SWE interviews, but still relevant. Most ML interviews include at least one coding round that may test data structures and algorithms. Being comfortable with LeetCode mediums (arrays, hashmaps, trees, basic graphs) is sufficient. You do not need to solve hard problems. Spend 30% of your coding practice on LeetCode and 70% on ML-specific implementations.
What programming language should I use in interviews?
Python is the standard for ML interviews. 95% of interviewers expect Python. Use NumPy for numerical operations, and know pandas basics for data manipulation. You will not need to use PyTorch or TensorFlow in the coding round (unless specifically asked), but you should be able to discuss them in the system design round.
How do I handle a question I do not know the answer to?
Be honest: "I'm not familiar with that specific concept, but let me reason through it based on what I know." Then work through the problem from first principles. Interviewers respect intellectual honesty and problem-solving ability far more than memorized answers. Never bluff — experienced interviewers will immediately detect it and it destroys your credibility for the rest of the interview.
Should I bring a portfolio or prepare a presentation?
Not for standard interviews. But have a well-maintained GitHub profile with 2–3 quality ML projects (clean code, good README, clear results). Some companies (especially startups) give take-home projects — treat these like production code with documentation, tests, and clear explanations of your approach and trade-offs.
How long should I wait between receiving an offer and responding?
Ask for 5–7 business days to evaluate the offer. Most companies will grant this. If you are waiting for other interviews to conclude, communicate this transparently: "I am finalizing interviews with [other companies] and expect to have all information by [date]. I want to make a fully informed decision." If they give you a hard deadline, you can ask for an extension, but respect their timeline.
What if I fail an interview? How long should I wait before re-applying?
Most companies have a cooldown period of 6–12 months before you can re-interview. Use this time productively: analyze what went wrong, strengthen weak areas, and build additional ML projects. Many successful ML engineers at top companies failed their first attempt. The key is to extract specific lessons from the failure and address them systematically.