Skip to content
Lubo Bali
BlueprintBook a call (opens Calendly in a new tab)

Available for new AI projects now.

I build AI that works on your real data, in production.

Senior Data & AI Platform Engineer. I built LuBot.ai on my own: a multi-tenant AI analytics SaaS that Neon (Databricks) featured for its one-database-per-customer design.

Book a free 30-min call (opens Calendly in a new tab)See my work
ActiveShipped today · 19,639 testsSee what I'm building
Lubo Bali, AI and data platform engineer, in a dark suit

Meet Lubo, and LuBot in action

Why most AI projects fail, and how I fix it.
Read the transcript

AI agents are everywhere right now. But most AI projects don't fail because of the AI. They fail because of the data behind it. An agent is only as good as the data you give it. And that's exactly what I fix.

I'm Lubo, and I've been in this field for over 15 years. For 12 years I worked with data, financial data with SQL first, then data engineering. For the last 3+ years, I've been building AI agents and MCP servers that connect AI to real business data. Those years taught me how to take care of people's data, and how to ask the right questions.

That's why I start with questions, not code. First, I understand your business and what you really want to get. Then I pick the technology, not the other way around.

So why me? First, I build everything myself, from start to finish. The data, the AI, and the app your team uses. I built my own AI platform, LuBot.ai, completely on my own, and Neon, part of Databricks, wrote an article about it.

Second, I build AI you can trust. In my systems, the numbers are calculated from your real data with real code, and the AI only explains them. So it can't make up a number, and the same question always gives you the same answer.

And third, I keep your AI bills low. In one project, 80% of the questions cost zero in AI fees.

You hire me per project. A free call, a simple contract with a written scope, an update every two weeks, and everything I build is 100% yours.

So tell me your problem, and I'll find a way to build a solution. Book a free call. Talk soon!

LuBot answering a real question about website traffic: chart, reasoning and raw SQL in seconds.

Featured by Neon (Databricks)

Inside LuBot's database-per-tenant architecture
  • 260,000+ lines of code, built solo
  • 12,000+ automated tests
  • 1 database per customer
  • 3 MCP directories list LuBot

What I do

Get your free AI blueprint

Tell me your problem the way you would on a call. In about a minute you get a first plan: how your work runs today, what it looks like with AI, the risks and a timeline.

See an example:
Where does your data live today?
How many people on your team?

Please describe it in general terms: no customer data or passwords. What you type is saved without your name, so I can learn what people need. Emails and phone numbers are removed. How I handle your data

Free · about 1 minute · one blueprint per visitor

Selected work

  • LuBot.ai

    Founder · AI analytics SaaS, built solo

    Ask questions about your website, files or stock portfolio in plain English and get charts, reasoning and SQL. Multi-tenant, one database per customer.

    • Python
    • FastAPI
    • PostgreSQL (Neon)
    • React
    • OpenRouter
    • DeepInfra
    • MCP
    • AWS S3

    Read the story Visit LuBot.ai Read the Neon article

    • Featured by Neon (Databricks) for its database-per-customer design
    • Runs inside Claude with 9 MCP tools, listed in 3 MCP directories
    • 260,000+ lines of code and 12,000+ automated tests

  • HelloPayments

    Senior Data & AI Platform Engineer · Mar⁠–⁠Sep 2026

    Lead engineer for a multi-product B2B SaaS for payment-processing sales organizations. Built DataSwim, an automated residual and commission engine that replaces manual Excel.

    • Python
    • FastAPI
    • PostgreSQL
    • Next.js
    • LiteLLM
    • NVIDIA NIM

    Read the story

    • Residual engine end to end: the pilot customer ran 6 months of real data through it, $31K+ in commissions
    • AI column mapper: a new payment processor needs zero code change
    • 3-tier LLM gateway: 80% of queries at $0 LLM cost

  • GraphGuard-AI

    Open source · ML for money-laundering detection

    Finds money laundering by the shape money moves in, not by any single transfer, across 5 million simulated bank transfers (IBM AMLworld).

    • Python
    • XGBoost
    • PyTorch Geometric
    • Polars
    • Optuna

    Read the story See the code on GitHub

    • 332 of the top 1,000 alerts are real laundering, against about 1 for a random queue
    • Surfaces 76% of complete laundering rings
    • Leakage rules enforced by CI; the graph model lost to XGBoost, and that is reported honestly

See all projects

What people say

Built with

A few of the tools I build with. Pick one to see a project I used it in.

How it works

  1. Step 1

    Free 30-min call

    You tell me the problem. I tell you honestly if AI is the right fit.

    Free, no obligation

  2. Step 2

    Contract and a written scope

    A simple contract, plus a written scope for each project: what I build, the steps and the timeline. Agreed before I start.

    Scope and timeline in writing

  3. Step 3

    Every 2 weeks: update + invoice

    Dates, work done and hours per task, tracked automatically. You pay for real work only, and any line you question gets sorted out first.

    Hours tracked automatically

  4. Step 4

    Handover

    Working code with docs and tests, so your team can run it without me.

    You own 100% of it

Fair terms, both ways

  • You own 100% of what I build: code, docs and designs.
  • Your data stays confidential and only in the systems you give me.
  • I finish what we agree on and keep you in the loop. You pay invoices on time.
  • Either side can stop with 14 days' notice. You keep all the work done so far.

Now Building LuBot.ai.

Available for new AI projects now

Ask me

Questions CEOs ask me

Pick a question to see my answer

Do you understand my business, or only code?

Lubo: Data first, then AI. I ran the finances of a 50-unit building for 12 years, so I start with how you make and lose money. Then I build.

✓ Proof: How I think →

Will it really save money? How will we know?

Lubo: We measure today first: hours, errors, time per task. Then we measure again after. One of my agents went from $12 and 8 minutes per question to $0.001 and 15 seconds.

✓ Proof: LangChain agents, 12,000× cheaper →

Will AI replace my people?

Lubo: No. I automate the copy-paste and the waiting. Your people keep the work where the human touch matters, and my agents ask a person before anything important.

✓ Proof: Sentinel-AI: it can never send or delete →

Can the AI make up numbers?

Lubo: Not in my systems. Real code calculates every number from your data. The AI only explains it, and every answer shows where it came from.

✓ Proof: How LuBot.ai does it →

Is my data safe?

Lubo: Encrypted, in your own cloud, never used to train AI. I design for it from day one. When a production server was hacked, I cleaned it up and locked it down myself.

✓ Proof: How I keep your data safe →

Will the AI bills get out of hand?

Lubo: No. Each question goes to the cheapest model that can answer it well. In one live project, 80% of questions cost nothing in AI fees.

✓ Proof: HelloPayments →

Will you tell me the truth, even when I don't like it?

Lubo: Yes. HelloPayments wanted one shared database for all their customers. I pushed for a separate database per customer, plus encryption. They listened. And if AI isn't the right fit for your problem, I'll tell you on the first call.

✓ Proof: HelloPayments →

What if you haven't used our exact tools?

Lubo: I'll tell you on the first call, then I learn them fast. AI tools change every month, and keeping up is part of my job. The patterns stay the same: clean data, tested code, secure by design.

✓ Proof: Check my fit for your stack →

Who owns what you build?

Lubo: You do, 100%. When the project is done, I hand over all the code and documents.

How fast will I see results?

Lubo: It depends on the project, but I build fast. You get a first small win early, and you see progress every week.

How much does it cost?

Lubo: A fixed price per project, agreed after a free 30-minute call. We start small, with one win you can measure.

Book a free 30-min call → (opens Calendly in a new tab)

My own answers, written by me.

Tell me your problem. I'll find a way to build the solution.

Available now

Book a free 30-min call (opens Calendly in a new tab)

No obligation. Pick a time that suits you.

Prefer email?data@lubobali.com