Wiki
Freelance Data and ML Careers
How two DataTalks.Club guests frame freelance data and ML careers through paid learning, public proof, lean MVPs, specialization, and client acquisition.
Related Wiki Pages
Freelance data and ML careers combine technical work with practice-building. Orell Garten shows the consultant path in DataTalks.Club. He moved from research and startup work into focused data-engineering services. Small useful deliveries helped him win trust (From Academic Research to Lean Data Consulting, 19:34-25:02).
Pastor Soto shows a learning-and-visibility path. He started with small remote data projects and learned under deadline pressure. Public ML projects and community work then helped him attract interviews and freelance opportunities (From Medicine to Machine Learning, 4:37-10:21 and 27:27-52:44).
Use Freelance Data Engineering and Consulting, Career Transitions in Data, Career Growth, and Job Search for the wider career context. The shared move isn’t “take any data gig.” It turns previous skills into credible client proof. It also chooses a narrow enough problem space and keeps delivery close to feedback.
Practice-Building Starts With Proof
Orell’s first freelance signal came through a contact from his startup period. Around 21:24 in From Academic Research to Lean Data Consulting, he says a previous contact returned with a small paid consulting request. The project mattered because it proved that his startup and data-platform skills had market value even after the company didn’t work out. He then focused on quality delivery. Networking, LinkedIn sharing, and referrals became part of the same practice (Orell Garten).
Pastor’s proof began smaller and earlier. Around 6:05 in From Medicine to Machine Learning, he describes signing up for Upwork. His first small payment came from helping with a statistics problem. Around 7:12-10:21, he explains how those projects pushed him from SPSS into Excel and R. Later client work demanded Python.
For a transition into machine learning or data engineering, the proof wasn’t a credential alone. It was evidence that he could learn under deadline and deliver something useful (Pastor Soto).
Learning by Doing Has Different Risk Profiles
Pastor frames early freelancing as a high-pressure learning environment. Around 9:39-10:21 in From Medicine to Machine Learning, he says people often asked him to take projects he didn’t yet know how to do. He learned quickly on the job. That worked for him because the projects created motivating deadlines. He was able to succeed in most of them.
The same passage also shows the intensity through early mornings and late nights. The work required constant skill acquisition.
Orell’s learning habit is more conservative once a client is involved. Around 46:43-53:03 in From Academic Research to Lean Data Consulting, he separates client value from personal experimentation. As a freelancer, he keeps new-technology experiments on his own time. He uses tutorials and small rebuilds to learn tools such as DuckDB. He also prefers something that works over a perfect solution that may never arrive.
In career growth, that means keeping a broad view of possible tools without making the client pay for unfocused exploration.
Client Acquisition Needs Visibility and Relationships
Both episodes treat client acquisition as its own skill, not as an automatic side effect of technical competence. Orell is explicit around 55:46 in From Academic Research to Lean Data Consulting: most people already have skills from full-time work. Acquiring clients is a different skill set for them.
Around 30:50-32:47, Orell describes mentioning that he’s self-employed when relevant. He also used recruiters for early freelance projects and relied on momentum once work started. Around 57:47, he adds that networking can be exhausting for introverts. Avoiding it still makes client acquisition harder (Orell Garten).
Pastor’s acquisition path moves from a marketplace to public reputation. Around 41:03-46:43 in From Medicine to Machine Learning, he says Upwork became harder after the pandemic. He opened LinkedIn and began posting course notes and ML explanations. Around 50:53-52:44, he says community participation and mentoring helped create new opportunities. Posts about concrete problems led people to ask for help on freelance and full-time projects.
For job search and freelance work, his episode treats visibility as a compounding asset rather than a one-time application tactic.
Lean MVP Delivery Comes Before Infrastructure
Orell’s freelance delivery starts with the smallest useful look at the data. Around 34:35-36:24 in From Academic Research to Lean Data Consulting, he defines his specialty as software-side data engineering for industrial clients. The work includes pipelines, data preparation, custom integration, and transformations for machines and formats that don’t arrive cleanly. Some clients know the target implementation. Others only know they have data and want analysis.
In the second case, a CSV can be enough for a first step if it exposes what’s possible and what’s broken.
The MVP workflow becomes concrete around 39:00-42:16 in the same Orell Garten episode. He starts by inspecting schemas and documenting the data. Then he pulls a small time slice locally and uses simple scripts to find a problem or insight before automating ingestion. Around 1:00:59, he adds that manual filtering or classification can be the fastest first iteration. It teaches edge cases that are hard to code for.
For readers coming from Data Engineering Portfolio Projects, a useful project should show judgment about source data and business value before showing a large platform.
Weekly Feedback Prevents Overengineering
Orell ties overengineering directly to premature infrastructure. Build before understanding the client problem and infrastructure may support too many imagined use cases (From Academic Research to Lean Data Consulting at 42:58).
Around 43:34-45:58, he describes regular client meetings as a forcing function for simple delivery. Weekly meetings can work, but the exact cadence is less important than the feedback loop. The freelancer shows what was done and discusses results, so complexity increases only when necessary.
That feedback loop is also a reputation mechanism. Around 44:50 in the same episode, Orell says he can’t disappear for six months and return with a perfect solution that may not be needed. It would be expensive and bad for his reputation. In freelance data work, the technical choice is therefore inseparable from the consulting relationship. Small demos, visible progress, and joint decisions keep the client close to the work.
Specialization Makes the Offer Legible
Orell’s offer is legible because it names the kind of data work he does. Around 34:35 in From Academic Research to Lean Data Consulting, he says he focuses on software-side data engineering rather than dashboarding or Power BI. Many of his clients work in industrial settings where machines, formats, and vendor variants require custom integration. Around 53:54, he also says data cleaning in those environments depends on domain knowledge and hours of client conversation. Changing data requires understanding what values mean for the business.
Pastor’s specialization is more identity-and-portfolio driven. Around 47:48-49:30 in From Medicine to Machine Learning, he explains that healthcare ML capstones helped make sense of his combined medical and data background. The examples used skin cancer and pneumonia data. The projects were dockerized and deployed on AWS. They were also reusable as proof when recruiters or project leads asked what he could do.
His route fits the broader career transitions route because prior domain experience becomes more useful when it’s attached to visible technical artifacts.
Public Learning Turns Work Into Market Memory
Pastor’s public-learning system is practical rather than decorative. Around 28:04-31:39 in From Medicine to Machine Learning, he says leaderboard participation pushed him to post weekly. It also pushed him to frame posts as explanations, not just “I’m learning” updates. Explaining topics such as ROC curves helped him appear as someone with professional insight. Around 32:50-34:32, recruiters reached out based on LinkedIn posts even though he wasn’t actively looking.
The workflow also improved learning. Around 44:15-46:43 in the same Pastor Soto episode, he describes using Notion or Google Docs notes. He turned one video or concept into several posts and double-checked material before publishing. That made notes, posting, recruiter visibility, and learning reinforcement part of one process. For freelance data and ML careers, public work is strongest when it shows a repeatable way of thinking, not just a finished project gallery.
Related Pages
These pages give broader context for the transition and delivery choices above.