803: How to Thrive in Your (Data Science) Career, with Daliana Liu

Episode Summary

Executive Summary: Daliana Liu explains why data science careers are less about titles and more about understanding day-to-day work, business impact, and personal fit. She shares how burnout and coaching led her to leave full-time DS for coaching/content, and offers advice on overcoming restrictive self-rules, imposter syndrome, and over-optimization while building influence, choosing a niche, and learning modern technical skills like SQL, vector databases, and GenAI APIs.

Main Topics: Career pivot from data scientist to coach and creator: Daliana describes leaving full-time roles at AWS and Predibase after years of balancing work with LinkedIn content and The Data Scientist Show, explaining that burnout and a deeper desire to help people pushed her toward coaching and educational content full time. Unhelpful career mindsets and self-imposed rules: She argues that people often invent rigid rules such as 'I must stay technical' or 'a podcast must continue weekly,' and that these beliefs trap them in roles or habits that no longer serve them. Imposter syndrome, confidence, and emotional 'wounds': Daliana discusses how childhood or early-career feedback can create lasting fear of speaking up, asking for promotions, or changing paths, and recommends journaling, brag logs, and self-awareness to build resilience. Finding a niche by understanding the real role: Rather than chasing titles, she advises researching actual job duties, company stage, team dynamics, and promotion criteria to discover the best-fit role—whether in product analytics, ML engineering, GenAI, or adjacent paths. Building influence and internal visibility: She emphasizes relationship-building, stakeholder management, internal demos, lunch-and-learns, and making work visible as key career accelerators, while warning against confusing networking with inauthentic politics. Technical skills that still matter in today’s market: Daliana highlights SQL, Pandas, software engineering basics, LeetCode-level problem solving, vector databases, and experience with GenAI APIs/RAG as practical skills increasingly expected across data roles. Content creation as career leverage, not necessarily a content career: She says public content can help build a personal brand, surface opportunities, and create artifacts others can find, but follower count alone does not drive promotions and can create workplace risks if handled carelessly.

Key Arguments: People should stop assuming there is one correct career path; many 'rules' are self-imposed rather than real constraints. Career dissatisfaction often comes from doing too much at once, not from a lack of ambition; burnout can be a signal to rethink priorities. The right question is not 'What course should I take?' but 'What does the day-to-day of that role actually look like?' Understanding the commercial context of a data project—cost, latency, legal risk, security, and stakeholder trust—is as important as model quality. Many data scientists overvalue technical perfection; in practice, good-enough solutions that meet business goals often matter more. To move up or pivot, it is usually best to build trust and transition internally rather than zigzag through unrelated intermediate roles. Influence is built through genuine relationships, visibility, and helping stakeholders succeed, not just by being technically excellent. Public content is most useful when it creates visible artifacts and learning, not when it is optimized for follower growth or vanity metrics. GenAI is important, but it will be woven into existing roles rather than creating entirely separate job titles in most companies. Confidence grows when professionals document wins, remember project goals, and recognize that success is often relative to a baseline, not perfection.

Data Points: LinkedIn followers: 280,000+ - Daliana’s audience size on LinkedIn, cited as evidence of her broad reach in career content. AWS tenure: 7 years - She spent seven years at AWS analyzing experiments, supporting launches, and building ML solutions. Content creation start year: 2020 - She began posting career/content material on LinkedIn around this year. Podcast duration as a grind: 3+ years - She said balancing a day job with weekly podcast/content production felt like having two jobs for over three years. Maven course enrollment: about 10 people - Her first live course launch on Maven attracted around ten participants. Course discount: 20% - She offered Super Data Science listeners a 20% discount code for her Maven course. Podcast guest audience feedback: April (last update mentioned) - A listener noted that The Data Scientist Show had not been updated since April. Internal project example: 60% to 70% - Used as an example of a meaningful but not perfect model improvement that can still count as success. Resume/project visibility: 1,000+ views - A post on her discontinued site ensemble.careers reached over a thousand views and led to a feature from Amazon Science. Website registrations: 30 people - ensemble.careers had roughly thirty registered users before she shut it down. Potential performance gain: up to 50% training costs, up to 40% inference costs - AWS sponsor message about Trainium and Inferentia chips for generative AI workloads. LLM use case: RAG more common than fine-tuning - She said many real-world company use cases are better served by retrieval-augmented generation than by fine-tuning.

Pivotal Quotes: "If you start a podcast, you have to keep going." — Daliana Liu: She used this as an example of a self-imposed rule that can keep people stuck in obligations they no longer want. "The goal of that project was to show them the possibility… I think it was already great." — Daliana Liu: Her manager reassured her that a demo project succeeded even if it never became a widely used production system. "If nobody can see it, that means it doesn't exist." — Daliana Liu: She was explaining why publishing projects, posts, or code online matters for career visibility and discoverability.

Implications: Data careers are becoming broader, more business-driven, and more hybrid. Professionals who combine technical depth with self-awareness, communication, and visible artifacts will adapt best to AI-era role changes and promotion pathways.