Building a robo-advisory engine for a neobank: our approach

Andrew Yakush — Mon, 06 Jul 2026 06:13:02 +0000

Robo-advisory sounds simple until you try building it inside a full banking platform. Here's how we're thinking about it at Y-tech Bank.

A useful robo-advisor isn't just "fill out a risk quiz, get a portfolio." That's a 2015 product. A modern one needs dynamic risk profiling that updates from behaviour instead of a one-time quiz, cross-domain context so investment decisions know about your banking data, awareness of life events so the portfolio adjusts when something changes, micro-investing built into the same flow, and explanations so users understand what the system is doing and why.

Building it inside a neobank gives you a real data advantage. Traditional robo-advisors like Betterment or Wealthify work with isolated investment data — they know your declared income and risk tolerance, and that's about it. A robo-advisor inside a bank can see your verified income from actual deposits, your monthly expense baseline from transaction history, your upcoming large expenses from recurring patterns, your emergency fund status, your business cash flow if you're an SME user, and tax signals like year-end expenses or self-employment patterns.

That changes recommendation quality a lot. If the system knows your rent is due in 4 days and your salary hasn't landed yet, it shouldn't be telling you to invest this week. If your income jumped 30% this month, it can bump up your contribution automatically.

Our architecture has the banking data layer feeding a single real-time financial model of the user, which feeds the investment engine — risk profiling, portfolio construction, rebalancing triggers, the micro-investing flow, and the explainability layer. The key decision: the investment engine reads from the shared user model. It doesn't keep its own separate profile.

Investment services in the EU fall under MiFID II, which adds requirements around suitability, best execution, cost disclosure and appropriateness testing. We're building MiFID II compliance in from the start, which means the explainability layer isn't just nice UX — it's a regulatory requirement.

One feature people underrate: micro-investing. Round-up investing sounds like a gimmick. The data says otherwise. Users who turn on micro-investing invest 3–5x more over a year than users who start investments manually, because removing friction works. Our flow: a transaction completes (say €4.60 coffee), the round-up gets calculated (€0.40), round-ups build up in a buffer, and at a threshold the user sets (say €5) the investment runs automatically. The user watches the portfolio grow alongside their spending, in the same session, same context.

Honest note on what we haven't solved: real-time rebalancing at scale is still a work in progress. The challenge isn't the algorithm — it's executing trades efficiently across thousands of accounts at once without slippage or latency problems. We're looking at execution partnerships for that part.

—

Y-tech Bank is pre-seed. We're building this stack and looking for engineers who want to work on hard fintech problems. ytechbank.com

How AI is actually changing banking (not just the hype)

Andrew Yakush — Fri, 03 Jul 2026 15:44:14 +0000

"AI-powered" might be the most overused phrase in fintech right now. Every bank claims it. For most of them it means they added a support chatbot and called it intelligence.

So let's talk about what real AI in a banking product looks like, and what we're building at Y-tech Bank.

Most current implementations fall into three shallow buckets.

Expense categorization. Your bank sees "McDonald's" and tags it "Food & Dining." Mildly useful. But this is rule-based sorting with an ML wrapper on top. It's automated, not intelligent.

Fraud detection. This one's actually valuable, and most banks have used ML here for years. But it's infrastructure, not a feature — users never see it working.

Chatbots. The most visible and the most annoying. "Hi, I'm Aria, your AI assistant. I can help you check your balance!" No, Aria. You can point me to the balance page.

Real AI in banking should predict things, act before you ask, and connect data across different areas. We're building three layers for that.

The first is personal finance intelligence. This goes past categorization into actual behaviour modelling. The system builds a live financial picture of each user — income patterns, spending rhythms, savings goals, risk tolerance — and uses it to forecast cash flow 30 days out, flag a possible overdraft before it happens, move spare cash into micro-investments automatically, and surface things like "you spent 40% more on subscriptions this month than your six-month average."

The second is NeuroOffice, our AI suite for small businesses. It's a set of agents inside the business banking layer — a marketer, copywriter, HR manager, accountant, lawyer, client manager. Each one has access to the business's real transaction data, so the accountant agent isn't working blind. It sees actual cash flow, real expenses, real revenue trends.

The third is investment intelligence. The robo-advisory layer takes both personal financial signals and market data to build portfolios that fit the user's current financial health, rebalance around life events like a salary change or a big purchase, and explain investment decisions in plain language.

The hard part isn't any single model. It's the shared data layer that lets all three talk to each other in real time. We build it with an event-driven architecture where financial events ripple across every AI surface, one shared user context that all the agents read from, personalization that keeps data inside the platform, and blockchain-backed transaction records for auditability.

The point of AI in banking isn't a better chatbot. It's a platform that actually understands your life and acts on it without you having to manage everything yourself.

That's what we're building. And we're hiring engineers who want to build it with us.

—

Y-tech Bank is a pre-seed fintech. The MVP is live. We're open to , investors and early users. ytechbank.com

DEV Community: Andrew Yakush

Building a robo-advisory engine for a neobank: our approach

How AI is actually changing banking (not just the hype)