Ep38: How to Land a Machine Learning Job Today
🎯 Summary
Podcast Episode Summary: Ep38: How to Land a Machine Learning Job Today
This episode of the Data Neighbor podcast features Umang Chaudhary, an ML Engineer at TikTok and former Software/ML Engineer at Amazon, discussing the current state of the ML job market, the impact of LLMs, and a structured approach to breaking into the field.
1. Focus Area
The primary focus is on navigating the Machine Learning Engineering (MLE) job market, addressing career anxieties driven by rapid AI advancements (especially LLMs), detailing the practical steps for interview preparation, and Umang’s personal journey transitioning from software engineering to ML within large tech companies.
2. Key Technical Insights
- LLMs in Production & Explainability: LLMs are increasingly integrated into production ML systems (e.g., risk detection at TikTok) not just for core detection but also for providing transparency and explainability (e.g., detailing why a user was suspended based on policy violations).
- The Unstructured Nature of ML Interviews: ML interviews are highly varied, covering end-to-end system design, scenario-based model selection, and ML coding. There is still a lack of a single, comprehensive resource for preparation, forcing candidates to learn through trial and error.
- Leveraging LLMs for Interview Prep: Modern LLMs (like Grok, which Umang specifically highlighted as comprehensive for ML system design) can serve as powerful, multi-turn interactive tutors for deep preparation, though many mentees still rely on traditional resources.
3. Business/Investment Angle
- Entrepreneurial Vision: Umang views his mentorship activities as a path toward entrepreneurship, aiming to build a sustainable business around ML career guidance alongside other ventures.
- Internal Mobility Advantage: Amazon’s lack of a mandatory lock-in period allowed Umang to transition from a general software engineering role to an ML role internally within months, highlighting the value of internal job marketplaces in large organizations.
- High Demand for AI/ML Roles: Despite fears of AI replacing jobs, the current theme among aspiring professionals is the urgency to secure an ML title now, as the structure and existence of these roles may change significantly in the next 3-5 years.
4. Notable Companies/People
- Umang Chaudhary: Guest speaker, current ML Engineer at TikTok (working on risk/malicious user detection), former MLE/SE at Amazon (transitioned from retail checkout to Prime Video forecasting).
- Jeff Hinton & Andrew Ng: Mentioned as early inspirations whose course videos sparked Umang’s interest in neural networks during his undergraduate studies.
- Amazon, TikTok, Meta, Microsoft: Mentioned in the context of internal mobility policies and hiring practices.
5. Future Implications
The conversation suggests a bifurcation in the ML field:
- Increased Automation: LLMs are rapidly boosting productivity for existing engineers and potentially automating simpler tasks, forcing new entrants to focus on deeper, specialized knowledge.
- Urgency to Enter: The window for securing traditional MLE roles might be closing or fundamentally changing, making the present moment critical for those seeking to enter the field before further structural shifts occur.
- Rise of AI Product Managers (AI PMs): There is a growing need for non-technical professionals (like Product Managers) to gain foundational AI knowledge to effectively lead ML-focused teams and products.
6. Target Audience
This episode is highly valuable for Aspiring Machine Learning Engineers, Career Changers (especially from Software Engineering or Data Analysis), and ML Mentors/Coaches. It is also relevant for Tech Professionals interested in Career Mobility within large tech firms.
Comprehensive Narrative Summary
The podcast opens by addressing the central anxiety plaguing aspiring ML professionals: Will AI automation eliminate MLE jobs before they even secure one? Host Shravya and Sean introduce Umang Chaudhary, who shares his journey to provide practical answers.
Umang detailed his path, starting from an undergraduate interest in neural networks, securing a research fellowship at NTU Singapore, and pursuing a Master’s in CS in the US. His entry into Big Tech at Amazon was unconventional; he was initially placed in a legacy web development role despite his ML focus. He leveraged Amazon’s internal mobility structure—noting the lack of a lock-in period compared to Meta or Microsoft—to aggressively interview internally. Within three months, he successfully transitioned to an ML role within Prime Video forecasting by treating the internal process like an external job search.
A significant discussion point was the difficulty of ML interview preparation. Umang stressed that initial failures stemmed from being a generalist; he learned the necessity of picking one domain and mastering the model details and trade-offs. He highlighted that ML interviews are a mix of system design, scenario questions, and coding, with limited consolidated public resources available, even today.
Umang now channels this experience into mentorship, driven by a desire to help others and build an entrepreneurial venture. He noted that the primary theme he hears from mentees is the existential fear regarding the future of MLE jobs. His advice is to act now: if one is passionate, this is the time to structure the learning, as the market structure is guaranteed to evolve.
Regarding LLM integration, Umang explained that at TikTok, LLMs are used in risk detection to enhance model performance and, critically, to generate transparent explanations for user suspensions. Personally, he uses tools like Gemini Pro and Grok extensively for coding assistance and deep dives into documentation. He strongly encourages mentees to use LLMs interactively for system design practice, noting that many fail to leverage these tools beyond simple, single-turn queries.
Finally
🏢 Companies Mentioned
đź’¬ Key Insights
"These are the four main areas that you have to break down your prep into, and all the interviews kind of fit all these puzzles only. Now, that includes later, LLM, that also includes your"
"in the end, it's about the numbers game. You have to target applying to 100 plus companies, try to get 10 plus tech screens, try to get 4 plus final loops, and then get at least one offer."
"right now the biggest advice is to stay patient. I see a lot of people lose hope after giving a few interviews and not getting selected. The unfortunate truth of 2025 and even 2026 will be that the market is tough."
"Meta recently said that they will be having a different kind of round where they will allow accessing LLM during the interview, but I think that still hasn't rolled out..."
"You are not getting asked very deeper PyTorch questions or very deeper TensorFlow questions, but you need to know the trade-offs between them. You need to know what the high-level intuition is for the majority of these things and understand where they can be used in the entire workflow, what are the trade-offs..."
"The critical piece of it is how do you actually interact with it? Meaning, it doesn't stop at one turn. It's by default a multi-turn system for you to get maximum value."