Julius AI: The AI Data Scientist for the Rest of Us

(5 minutes)

Hi 👋, this is the Today in AI Newsletter: The weekly newsletter bringing you one step closer to building your own startup. 

We analyze a cool, industry-shaping AI startup every week, with a full breakdown of what they do, how they make money, how much they’ve raised, and the opportunity ahead.

Let’s get to the good stuff in this email:

• 📊This startup is building an AI tool that can read and visualize your data with natural language.

• 📈They’ve crossed 2,000,000+ users, with 10,000,000+ visualizations and 4,000,000+ lines of analysis code executed every 24 hours.

• 💰They just raised a $10M seed led by Bessemer Venture Partners, with Y Combinator, Horizon VC & more.

So what’s the startup and who are the founders behind it? Here’s the story of Julius AI 📈

Julius AI calls itself an AI data scientist/analyst. The mission: democratize data analysis so anyone - CFOs, operators, marketers, researchers, students, can turn messy datasets into clear answers.

How it works (in plain English 🗣️→🤖→📊)

1. Natural Language Prompting 🗣️: Users interact with Julius by talking to it in plain English - similar to talking to a human analyst.

2. Tool Use and Code Generation 🧑‍💻: Julius uses leading models (including GPT-5, Claude, Gemini) to convert them into executable code.

3. Code Execution and Sandboxing 📥: The AI executes this code (mostly Python, with R launching soon) in a secure, remote code sandbox or Virtual Machine (VM) in the cloud. 

This infrastructure is provisioned individually for each user, providing dedicated resources (e.g., up to 8GB or 32GB of RAM).

4. Self-Correcting Loop 🧠 (Advanced Reasoning): Julius uses an iterative, self-correcting loop, enabling it to handle complex tasks. 

The AI can devise a plan, execute the code, look at any errors thrown (e.g., syntax, dependency issues), and then rewrite or modify the code until the task is successfully completed. 

5. Data Handling 🔌: Julius can work with various files (Excel, SPSS, PDFs, images) and connect directly to data sources like Postgres, Google Drive, and SharePoint. 

When querying databases, Julius first reasons through the schemas, estimates the row count to avoid expensive queries, and then samples data.

Why isn’t this a “GPT-wrapper”? While general models (ChatGPT, Claude, Gemini) can “do analysis,” Julius specializes in real-world data. 📊

That focus yields better accuracy, and a product experience designed for analysis, not just chat. 🎯

Vision: dashboards, report generation, and memory so Julius proactively surfaces insights before you ask—like having a data scientist on standby 24/7.

Backstory 👀

Rahul Sonwalkar grew up with no early exposure to tech. 

After watching The Social Network around age 11–12, he realized software could reach millions from a single laptop. 

In school, he had an email service for 12 friends after forgetting his Gmail password.

He earned a full-ride to a Texas state school, felt “starved for ambition,” and optimized coursework (heavy exam-weight classes) to free time for projects. 

In college he built Waterview, a managed service for hackathons (login, DB, backend) that reached 17,000 developers - and taught him a crucial lesson: retention beats sign-ups.

After graduating (2020), Rahul joined Facebook (misinformation) and Uber (rider pricing). 

This is when he realized “one no can kill your idea”, which pushed him to quit and build full-time.

Rahul then joined Y Combinator in 2022 with a logistics routing /visibility platform that shippers loved - but 60-year-old truckers wouldn’t share 24/7 location.

Rahul “Ligma” 🤪

Around this time, Elon Musk acquired Twitter (Now X). Rahul decided to pose as a laid-off Twitter engineer "Rahul Ligma". The stunt went pretty viral and gained him "some notoriety".

His YC partner praised him for demonstrating "great founder energy" because he was seen as someone who was "hacking systems, especially social systems". 💻

Even today, Rahul says that he gets recognized in SF quite a lot on the street, and people sometimes ask for selfies. He actively leverages this recognition by wearing a Julius T-shirt, it’s "free marketing for Julius". 🤷

The Hustle 🤑

Pre-ChatGPT, Rahul was obsessed with GPT-3 and Copilot. He realized his heart (and the future) was in AI code-writing.

Early AI forays included text-to-SQL, ExcelCopilot.com (which drew a C&D from Microsoft), and Hoops GPT for NBA analysis - timing and SQL’s limits hurt. 

The breakthrough: ditch SQL for Python. He incorporated Caesar Labs, and Julius launched July 12, 2023 (yes - Julius Caesar’s birthday). 

They even had paying users Day 1, but didn’t notice for two days because Stripe was in test mode. 😅

They started gaining traction on the GPT Plugin Store, but it shut down overnight - growth evaporated. ❌

Julius rebuilt on shareability (word-of-mouth/SEO), doubled down on infra, and leaned into the reality: people might call you a “GPT wrapper”, but the real moat is the product layer - the experience, speed, reliability, and ecosystem.

Rahul’s hiring mantra is “slope + agency” - multipliers who make everyone better. 

Two of the first four hires (then interns) dropped out of college to join.

Julius now uses the best models available and focuses its energy on UX + infra where users actually feel the difference. 🖥️

Engineers do support. Rahul personally answered 5,100 tickets, keeping the feedback loop insanely tight. 🔄

Julius wants to be the default layer where business gets analyzed. Replace Excel, Tableau, “all of that.” 

Stats 📊

📈Today, Julius AI has over 2 million users, up from 1 million just 6 months ago, and up from 0, just 18 months ago. 

📊The tool has helped operators, students and researchers create over 10 million data visualizations and helps them execute 4 million+ lines of analysis code every 24 hours.

Their underlying infrastructure serves over 1 million Jupyter notebooks - even though most Julius users are non-technical and don’t know what a Jupyter notebook even is.

They currently have a team of over 15 people and growing 🧑‍🤝‍🧑. They’re currently hiring: 

• Senior Software Engineer - Backend

• Senior Software Engineer - Frontend

• Senior Software Engineer - Infrastructure

• Product Designer

• Growth Marketing Manager

Julius AI is even a requirement in a Harvard Business School course for 900+ students; 86,000 Julius messages sent. 📚

With Julius, the cost of getting an insight from data is already down ~99% in 2 years - unlocking analysis for every desk.

💰In July 2025, Julius AI raised a $10M seed led by Bessemer with participation from YC, Horizon VC, 8VC, AI Grant, and angels including Guillermo Rauch (CEO of Vercel), Amjad Masad (CEO of Replit), Aravind Srinivas (CEO of Perplexity), Akshay Kothari (Co-founder of Notion).

The Opportunity 🚀

Every company sits on a goldmine of data that goes unused because analysis requires specialized skills. Julius flips that: the analyst becomes software. 🧑‍💻

1. Julius replaces day-to-day work in Excel/Tableau with plain-English analysis that writes code for you.
   

2. The tool serves the massive population of non-technical knowledge workers: Finance, Sales/Marketing/Ops, Research/Academia, Education who need answers, not tooling.
   

3. With the 99% cost collapse of data analysis, the TAM isn’t just data science - it’s every role that asks questions of data.

How to get the most out of Julius 💡

1. Just talk to it: Ask in plain English: “compute returns, run a regression, forecast sales.”
   

2. Set time boundaries 🗓️: “Last 3 months,” “Q4” → partition pruning saves ~80–90% of scanned data.
   

3. Be specific about outputs 🎯: “Return purchase frequency + AOV” vs. “analyze our customers” → cut costs ~70–95%.
   

4. Smart sampling 🔍: “Use a 20% sample” to find patterns fast.
   

5. Layer your questions 🪜: “Top 5 categories → drill into #1.”
   

6. Reuse summaries ♻️: Query monthly_revenue_summary instead of re-scanning millions of rows.

Before you go 👋

Thank you for reading. Know someone who'd love this newsletter? Share the love. 🌟

🚀 Building something in AI? Get featured in front of 90,000+ AI founders, operators, and investors. 👉 [email protected]