GDG-Yangon-DevFest-Navigating Machine Learning Career Paths

Transcript

2. • Ph.D. (Computer Science) Graduate, Faculty of Science, Prince of Songkla University • Machine Learning Google Developer Expert (Dec 2023-Now) • Machine Learning Researcher with an interest in medical NLP, sentiment analysis, social media analysis, and information extraction • 10 years of working experience in Natural Language Processing projects focused on sentiment analysis, social media monitoring, and machine learning applications • Published scientific articles in international peer-reviewed journals and conferences • Current Affiliations - Lecturer (KMUTT, Bangkok, Thailand)/Adjunct Professor (Parami University, Washington D.C, USA) GDG Yangon/Online
3. ML companies Job Demands and Job Security ML people ML career advices AMA Topics
4. Information Acquisition ★ Own experiences ★ Talk to other experts from SEA region ★ Medium articles ★ Tech news Yangon
5. Types of Machine Learning Companies GDG Yangon/Online
6. Product Companies Service Companies Enterprise in-house teams Universities Research Institutions
7. Product Companies Focus on building AI-driven products or platforms. Examples: Google (Search, Bard), OpenAI (ChatGPT). Opportunities: Product innovation, scalable solutions
8. Product Companies Pros • Impact millions through scalable solutions. • Well-funded projects and cutting-edge tools. • Exposure to full AI lifecycle: R&D to deployment. Cons High-pressure, competitive environment. Focus on product constraints may limit exploratory research. Unstable job security in start-up companies due to failing customer acquisition
9. Service Companies Provide tailored AI solutions to clients. Examples: Deloitte, Accenture, Sertis Opportunities: Client-facing roles, diverse problem-solving scenarios.
10. Service Companies Pros • Diverse projects across industries. • Opportunities to engage directly with clients. • Develop adaptable problem-solving skills. Cons • May lack depth in specific domains. • Often requires extensive travel or fast-paced work cycles. • Unstable job security in start-up companies
11. Enterprise In-house teams Develop AI solutions for internal business processes. Examples: Walmart AI Labs, Netflix (personalization algorithms). Opportunities: Domain-specific AI applications, long-term projects.
12. Enterprise In-house teams Pros • Stable, long-term focus on domain-specific solutions. • Integration with broader business strategies. • Opportunities for cross-department collaboration. • Good for job security Cons • Slower pace of innovation compared to AI startups. • Limited exposure to diverse problems.
13. Universities Focus on education and cutting-edge academic research. Examples: Stanford AI Lab, CMU Machine Learning Department. Opportunities: Research, teaching, student mentorship, publications.
14. Universities Pros • Intellectual freedom and opportunity to explore cutting-edge ideas. • Influence through teaching and mentoring. • Potential for collaborative grants and projects. • Good for job security Cons • Limited funding and slower implementation cycles. • Balancing teaching, research, and administrative duties.
15. Research Institutions • Dedicated to foundational and applied AI research. • Examples: Allen Institute for AI, DeepMind • Opportunities: Long-term innovation, collaborative projects, open science.
16. Research Institutions Pros • Focus on long-term, impactful AI advancements. • Access to extensive computational resources and collaborations. • Publish in high-impact venues. Cons • May lack immediate industry applications. • Competitive environment for funding and recognition. • Job security strongly depends on the number of publications and grants
17. 1.Reputation 2.Skill Development 3.Career Growth 4.Financial Security 5.Team Culture Consider Your Career Goals Product Companies: Innovation & impact at scale. Service Companies: Variety & dynamic problem-solving. Enterprise Teams: Industry-specific AI roles. Universities & Research: Knowledge generation & dissemination.
18. Job Security Product Companies Moderate to High ◦ Stability depends on the company's market performance and profitability. ◦ Risk of layoffs during economic downturns or product de-prioritization. ◦ Best for: Those willing to take some risk for innovation and career growth. Service Companies Moderate ◦ Projects are tied to client contracts, which can lead to fluctuations in demand. ◦ Job security improves with seniority and client relationships. ◦ Best for: Professionals who can adapt to changing roles and environments.
19. Job Security Enterprise In-House Team High ◦ Positions are tied to the company’s core operations rather than market trends. ◦ Lower risk of layoffs compared to startups or product companies. ◦ Best for: Those seeking a stable, domain-specific role. Universities Very High (with tenure), Moderate (without tenure) • Tenured faculty enjoy exceptional job security. • Non-tenured roles or adjunct positions may have limited stability • Best for: Those committed to academia and long-term career growth.
20. Job Security Research Institutions Moderate to High ◦ Depends on funding availability and research grants. ◦ Secure for well-established researchers or permanent positions. ◦ Best for: Researchers with proven track records and funding acquisition skills.
21. Transition Between AI Organizations GDG Yangon/Online
22. Moving to Product Companies • From Service Companies: ◦ Highlight product development experience from client projects. ◦ Showcase understanding of scalable and user-focused AI solutions. • From Academia/Research: ◦ Emphasize applied research with a tangible impact. ◦ Build a portfolio showcasing deployed models or prototypes. Moving to Service Companies • From Product Companies: ◦ Highlight versatility in solving diverse problems. ◦ Demonstrate adaptability to new industries and client-focused collaboration. • From Academia/Research: ◦ Focus on practical problem-solving and interdisciplinary skills. ◦ Show ability to manage multiple projects simultaneously.
23. Moving to Enterprise In-House Teams • From Product Companies: ◦ Emphasize experience with domain-specific AI applications. ◦ Demonstrate understanding of aligning AI solutions with business needs. • From Academia/Research: ◦ Showcase deep technical knowledge applied to enterprise challenges. ◦ Highlight ability to work on long-term, stable projects. Moving to Universities • From Industry (Product/Service/Enterprise): ◦ Build academic credentials: publish papers, attend conferences. ◦ Highlight mentoring and teaching experience within teams. • From Research Institutions: ◦ Leverage existing publications and funding acquisition experience. ◦ Demonstrate commitment to teaching and student development.
24. Moving to Research Institutions • From Academia: ◦ Leverage specialization and academic network. ◦ Highlight impactful and collaborative research projects. • From Industry (Product/Service/Enterprise): ◦ Show ability to translate practical challenges into research opportunities. ◦ Build credibility by publishing in peer-reviewed journals or presenting at conferences.
25. Tech Events Seminars Public Speaking Portfolio Tips for Smooth Transitions Skill Adaptation: Tailor your skills to align with the target organization's priorities. Networking: Leverage connections within the target industry or academic field. Personal Branding: Update your portfolio, LinkedIn, and resume to showcase relevant achievements. Upskilling: Take certifications or online courses to fill skill gaps (e.g., teaching skills for academia). Mentorship: Seek advice from those who have transitioned successfully.
26. Mock Up Case Study: Transition from Academia to a Product Company Background • Name: Dr. Maya Patel • Previous Role: Assistant Professor in Computer Science at a mid-tier university. • Expertise: NLP research, with publications • Goal: Transition to a role in a product company to work on real-world NLP applications. Challenges • Lack of Industry Experience: a. No prior exposure to working in fast-paced industry environments. • Limited Exposure to Scalable Systems: a. Research was focused on theoretical advancements rather than deployment. • Networking: a. Few connections in the tech industry outside of academia. Strategies for Transition 1. Skill Enhancement: ◦ Took an online course in ML Ops to understand model deployment and monitoring. ◦ Completed Kaggle competitions to showcase hands-on data processing skills. 2. Portfolio Development: ◦ Open-sourced research code on GitHub with detailed documentation. ◦ Built a prototype of a sentiment analysis app using a pre-trained NLP model. 3. Networking: ◦ Attended industry-focused conferences ◦ Reached out to former academic peers now working in the tech industry. 4. Targeted Applications: ◦ Applied to companies which value R&D skill, emphasizing transferable skills. ◦ Tailored her resume to highlight real-world impact of her research (e.g., improving chatbot algorithms). 5. Interview Preparation: ◦ Practiced coding interviews on LeetCode. ◦ Researched company-specific use cases for NLP to align her pitch.
27. Machine Learning People Understanding Career Pathways in AI GDG Yangon/Online
28. https://fullstackdeeplearning.com/course/2022/lecture-8-teams-and-pm/#common-roles
29. Read JDs carefully Job titles are interchangeable in product/service/enterprise in-house team
30. Roles The MLOps is primarily a software engineering role, which often comes from a standard software engineering pipeline. The ML Engineer requires a rare mix of ML and Software Engineering skills. This person is either an engineer with significant self-teaching OR a science/engineering Ph.D. who works as a traditional software engineer after graduate school. The ML Researcher is an ML expert who usually has an MS or Ph.D. degree in Computer Science or Statistics or finishes an industrial fellowship program. The ML Product Manager is just like a traditional Product Manager but with a deep knowledge of the ML development process and mindset. The Data Scientist role constitutes a wide range of backgrounds, from undergraduate to Ph.D. students. https://fullstackdeeplearning.com/course/2022/lecture-8-teams-and-pm/#common-roles
31. Universities Roles: • Lecturer/Professor: Teach and guide students in AI. • Researcher: Publish in peer-reviewed journals; mentor Ph.D. students. • Postdoctoral Fellow: Conduct cutting-edge research. Key Characteristics: • Emphasis on academic exploration and knowledge dissemination. • Flexibility in choosing research topics. Growth Path: • Lecturer → Assistant Professor → Associate Professor → Full Professor
32. Different People leads to Different Metrics GDG Yangon/Online
33. Example - Food ordering app •ML engineers – want a model that recommends restaurants that users will most likely order from, and believe they can do so by using a more complex model with more data. •Sales team – wants a model that recommends the more expensive restaurants since these restaurants bring in more service fees. •Product team – wants a model that return the recommended restaurants in less than 100 milliseconds. •ML platform team – want to hold off on model updates to prioritize improving the ML platform. •Manager – wants to maximize the margin, and one way to achieve this might be to let go of the ML team
34. My experience Problem – Hate Speech Detection from Social Media (Burmese, Sinhala, and Tamil languages) ML Solution – Classification Problem (Binary – Hate speech/No-Hate Speech) / Regression (A numerical score for hate speech intensity) Challenges • How to define hate speech?? (Took more than 4 months to get data labelling guideline) • Binary classification is not enough • Not all countries have the same hate speech scenarios (need political context knowledge) • hierarchical classification or multiclass classification are requested by the clients • A new category will be added in the future which leads to re-training the model from the beginning
35. Business Metrics and ML Metrics ML objective Accuracy, F1 Latency Business objective To maximize profits of shareholders Increase sales Cutting costs High customer satisfaction Increase time spent Chip Huyen, Designing Machine Learning Systems (Chapter 2)
36. 1.CV 2.Internal network 3.Stand Out 4.Business Value 5.Communication Job Hunting • LinkedIn • Facebook Group • Advisors’ Recommendations (Academics) • Interview preparation ◦ Leetcode ◦ Research projects (STAR method) • Others ◦ Salary ◦ Visa ◦ Relocation support
37. Set clear career goals Networking Portfolio & Presentation Preparation is the key. Failure is a common Be true to yourself My Christmas present
38. Thank you Q & A