In the author's Spotlight series, the TDS editor chats with community members about the career paths, writing and sources of inspiration for data science and AI. Today we are excited to share our conversation Claudia NG.
Claudia is an AI entrepreneur and data scientist with over six years of experience building production machine learning models with FinTech. She came in second place and won $10,000 in the 2024 Web3 Credit Scoring ML Competition.
You recently won $10,000 in the machine learning competition – Congratulations! What was the biggest lesson you took away from that experience, and how did it shape your approach to real-world ML problems?
My biggest lesson was realizing that domain expertise is more important than algorithm complexity. This is a competition for Web3 Credit Scoring ML and I have never dealt with blockchain data or neural networks for credit scoring, but my over 6 years at FinTech has given me the intuition of the business to treat this as a standard credit risk issue. This perspective has proven to be more valuable than the specialization of deep learning.
This experience fundamentally changed how we approach ML problems in two ways.
First, I learned that what was shipped is better than perfection. I only spent 10 hours competing and submitted a “MVP” approach rather than overengineering it. This applies directly to industry work. A decent model running in production offers more value than the highly optimized models found in the Jupyter notes.
Secondly, I discovered that most barriers are mental and not technical. I hardly got into it because I didn't know Web3 or felt like a “competitor,” but looking back, I was thinking about it again. I'm still working on applying this lesson more broadly, but I've changed the way I evaluate opportunities. I now focus on whether I understand the core issue, whether it excites me, and I believe I can understand it as I go.
Your career path spans business, public policy, machine learning and now AI consultants. What motivates the transition from corporate technology to the AI freelance world and what excites you most in this new chapter? What challenges and what clients are you most excited about working with?
The transition to independent work was something I really wanted to own and grow. The company's role will create valuable systems that will live longer than tenure, but you will not be able to take them home or earn continuous credits for success. Winning this competition shows that you have the skills to create your own solutions rather than contributing to someone else's vision. I have learned valuable skills in corporate roles and look forward to applying them to issues that I care deeply about.
I am pursuing this in two main passes. A consulting project that leverages data science and machine learning expertise and building AI language learning products. While consulting work provides immediate revenue and continues to lead to real business issues, the language products represent my long-term vision. I'm learning to build in public places and share my journey through my newsletter.
As a polyglot that speaks nine languages, I thought deeply about not only when learning foreign languages, but also the challenge of achieving not only textbook knowledge, but also the flow of conversation. I am developing an AI language learning partner that helps people practice real-world scenarios and cultural contexts.
What excites me the most is the technical challenges of building AI solutions that take the nuances of cultural context and conversation. The consulting side is energized by working with companies that want to solve real problems rather than implementing AI to use it. Whether you're working on risk models or streamlining information searches, I love projects where domain expertise intersects with practical AI.
Many companies are keen to “do something with AI,” but they don't always know where to start. What is the typical process for supporting the scope of new clients and prioritizing initial AI initiatives?
Rather than leading in AI solutions, we take the problem first approach. Too many companies want to “do something with AI” without identifying the specific business problem they are trying to solve. This usually leads to an impressive demonstration that doesn't move the needle.
My typical process follows three steps:
First, we will focus on diagnosing the problem. Identify specific problems with measurable effects. For example, I have recently been working with clients in the restaurant space to slow revenue growth. Instead of jumping to “AI-powered solutions,” I looked at customer review data to identify patterns. For example, we showed which menu items caused complaints, which service elements generated positive feedback, and which operational issues were most frequently displayed. This data-driven diagnostics led to specific recommendations rather than a general AI implementation.
Second, pre-defined success. I argue for quantifiable metrics such as time savings, quality improvements, and increased revenue. If it cannot be measured, it cannot prove that it worked. This prevents scope creep and not only builds cool technology, but also solves real problems.
Third, run a viable solution and align it with the best solution. It could be a visualization dashboard, a rag system, or it could be a prediction feature. AI isn't always the answer, but when it is, we know exactly why we are using it and how we are successful.
This approach has resulted in positive results. Clients usually see better decision-making speeds and clearer data insights. While I was building independent practices, focusing on real problems rather than AI buzzwords was key to client satisfaction and repetitive engagement.
He taught aspiring data scientists. What are the common pitfalls seen among people trying to break into the field?
The biggest pitfall I see is trying to learn everything, not focus on one role. Many people, including myself, feel early on that they need to take all AI courses and master every concept before they “qualify.”
The reality is that data science has a very different role. From product data scientists running A/B testing to ML engineers deploying models in production. You don't have to be an expert on everything.
My advice: Choose your lane first. Focus on figuring out which roles will excite you the most and hone your core skills. I personally moved from an analyst to an ML engineer, researching machine learning vigorously and taking on real projects (here can read my transition stories). I leveraged domain expertise on credit and fraud risk and applied this to make engineering and business impact calculations work.
The key is to apply these skills to real problems, not to get stuck in tutorial hell. I've always seen this pattern through my newsletter and mentoring. Those who break through are those who start building, even if they're not ready.
The landscape of the role of AI continues to evolve. How should newcomers focus on ML engineering, data analytics, LLMS, or something else?
Start with your current skill set and what you are interested in. I have worked in a variety of roles (analysts, data scientists, ML engineers), each of which has provided valuable and transferable skills.
Here's how I approach the decision:
If you are coming from a business background: The role of a product data scientist is often the easiest entry point. Focus on SQL, A/B testing, and data visualization skills. These roles often place more emphasis on business intuition than deep technical skills.
If you have programming experience: Consider ML engineering or AI engineering. There is high demand and can be built on existing software development skills.
If depicted in the infrastructure: MLOPS engineering is in great demand, especially as many companies deploy ML and AI models at large scale.
Although the landscape continues to evolve, as mentioned above, domain expertise is more important than following the latest trends. I won that ML competition. Because I knew the most fancy algorithms, but because I understood the fundamentals of credit risk.
Continue your technical skills with a focus on solving real problems in domains you understand. For more information about the various roles, I have written about five different data science career paths here.
What are the topics of AI or data science that you think more people should be writing about, or what are some of the trends you're looking at now?
I was blown away by the speed and quality of Text-to-Speech (TTS) technology in mimicking real conversation patterns and tones. I think more people should write about TTS technology for the sake of endangered language preservation.
As a polyglot with a passion for intercultural understanding, I am fascinated by the ways in which AI can help prevent language from disappearing completely. Most TTS developments focus on major languages with large datasets, but there are over 7,000 languages worldwide, many at risk of extinction.
What excites me is the possibility that AI will create speech synthesis for a language with only hundreds of speakers remaining. This is a technology that provides humanity and cultural preservation at its best! When language dies we lose our unique thinking about the world, specific knowledge systems, and cultural memory that cannot be translated.
A trend I often see is that transfer learning and voice cloning make this technically feasible. In particular, we have reached a point where we may only need a few hours, rather than thousands of hours of audio data, to create quality TTs for new languages, using existing multilingual models. Although this technology raises valid concerns about misuse, applications such as language preservation demonstrate how these features can be used responsibly for cultural benefits.
As I continued to develop language learning products and build consulting practices, I was constantly reminded that the most interesting AI applications often come from combining technical capabilities with deep domain understanding. Whether you're building machine learning models or cultural communication tools, magic happens at intersections.
You can follow her on TDS, Substack, or LinkedIn to learn more about Claudia's work and stay up to date with the latest articles.
