Daily Dose of Data Science: From a Cancelled Master's to 200K Newsletter Subscribers

🌏 中文版

Avi Chawla spent four years turning a solo Substack newsletter into a 200K+ subscriber AI learning platform. It didn't start with a grand plan — it started with a "what now?" after losing a graduate school admission.

How It Started: An Accidental Pivot

Avi Chawla graduated from IIT Varanasi (Indian Institute of Technology) and spent roughly two years as an AI Engineer at Mastercard. In 2022, he was admitted to the University of Maryland, College Park's AI research master's program — one that accepts only 10–15 students per year.

Two months before departure, family circumstances forced him to cancel.

Rather than return to Mastercard, he used the gap to reorganize his data science knowledge. Inspired by knowledge-creator Ali Abdaal, he launched Daily Dose of Data Science on Substack.

As he wrote in his one-year retrospective:

"The newsletter became unexpectedly rewarding and deeply meaningful during a challenging personal period."

Growth Timeline

Exceeding a full-time salary in five months — while each 100–150 word post took 3–4 hours of research, experimentation, and drafting. India's presumptive taxation scheme helped: freelancers pay taxes on only 50% of income.

Content Strategy: Curator, Not Researcher

The core format is one 100–150 word post per day plus a visual explainer — a two-minute read.

The key insight: Avi doesn't invent algorithms or run experiments. He distills one concept from a paper, official doc, or GitHub repo into a single visual and a few sentences of explanation. He's fundamentally a technical curator — and AI/ML is a field where dozens of arXiv papers land daily, GitHub trending surfaces new frameworks constantly, and major companies keep shipping new models. The bottleneck is never "what to write about" but "what to pick."

Early on, he funneled readers from his existing Medium / Towards Data Science audience to Substack. Later, his companion GitHub code repository hit 35K+ stars and became a major discovery channel on its own.

There was one deliberate content pivot: from covering trending tools to explaining core concepts — PCA, GLMs, Bayesian optimization. These topics don't expire, can be revisited from different angles, and form better backbones for paid courses.

Business Model: Free Newsletter to Paid Learning Platform

His monetization followed a classic creator economy ladder:

1. Free daily newsletter (2022 onward): Build trust and an audience base
2. Substack paid tier: Free version shows summaries only; paid unlocks full articles
3. Lab community (late 2023): Small-cohort live sessions plus forums, capped at 120 members — a high-ticket product
4. Free guidebooks: MCP, Agents, AI Engineering, and Data Science — 500K+ total downloads, serving as top-of-funnel lead magnets
5. Crash courses (8 total): Structured paid courses on RAG, agents, MCP, LLM fine-tuning, MLOps, and more
6. Standalone membership site: Migrated from Substack to dailydoseofds.com, consolidating all paid content

He hasn't disclosed specific revenue figures. But given Substack tech newsletters' typical 5–10% free-to-paid conversion rate, a 200K subscriber base, plus course and community revenue, annual income is likely in the six-figure USD range.

How to Ship Daily for Four Years Without Burning Out

The most common question: how do you keep producing daily without running dry? From his public posts, a few patterns emerge:

Short format lowers the physical bar. 150 words plus one visual — not a 3,000-word essay. This makes "one per day" physically sustainable.

AI/ML has a natural information surplus. Dozens of new arXiv papers daily, new GitHub repos, new model releases from major companies. A curator in this field never lacks raw material.

Learning-in-public flywheel. Writing forces you to actually understand a fuzzy concept; reader questions expose blind spots; blind spots become next week's content. The loop is self-sustaining.

Concept content doesn't expire. Tools get replaced, but PCA's math doesn't change. After pivoting to core concepts, the same topic can be rewritten from different angles — the source pool is effectively infinite.

He stopped going solo. A year in, he publicly sought a co-founder — "no breaks, working day in and out" and "extremely tiring and stressful" were his own words. It wasn't until Akshay Pachaar formally joined in 2025 that the production burden was truly shared.

Three Lessons He Shared

Imposter syndrome is real. Working independently with no colleague feedback, he maintained a document of positive reader messages to read during low points. Simple but effective — he's mentioned this practice repeatedly across public posts.

Solo founders hit a ceiling. Seeking a co-founder after just one year shows that "one person achieving positive revenue" and "one person sustaining it long-term" are different problems.

Chasing tools is ephemeral; teaching concepts lasts. This pivot simultaneously solved content sustainability and the business model problem — concept-based content is better suited for structured courses with higher price points than individual article subscriptions.

The Bigger Picture

Daily Dose of Data Science's success rests on conditions worth acknowledging: Avi had an existing writing audience on Medium (not a cold start), AI/ML happens to be the most information-oversupplied vertical there is (naturally suited to daily curation), and India's tax structure and cost of living make the financial bar for full-time newsletter work far lower than in Silicon Valley.

But the real takeaway is the format choice: 150 words plus one visual. This deliberately compressed format simultaneously solves the reader-side problem (low barrier, high completion rate) and the creator-side problem (physically sustainable output). It's the physical foundation that makes the entire flywheel spin.

For anyone considering technical content creation, the biggest lesson from this case isn't "you too can earn six figures." It's this: with the right format and domain, curation is more sustainable than original creation — provided you can endure a year of solitude.

References

• Daily Dose of Data Science — Official Site
• Daily Dose of Data Science — A Year in Review and What's Next (Avi Chawla's one-year retrospective)
• Avi Chawla — 10K Subscriber Milestone (LinkedIn)
• Avi Chawla — Seeking a Co-founder (LinkedIn)
• ML Spring Merging with Daily Dose of Data Science
• ChawlaAvi/Daily-Dose-of-Data-Science — GitHub