BRCI — Banca Română de Credite și Investiții × Carol Calin
Confidential proposal · prepared for BRCI

“We want to grow together.”
Let's also grow the people who use AI.

BRCI grows without a large branch network — online business accounts, BRCI iBanking and Mobile, instant payments, deposits and government securities. I'd start where it matters most at your scale: people. Hands-on sessions so the “Ask a specialist” team uses AI with confidence, then agents for onboarding and compliance & fraud triage.

For  The executive management, BRCIBy  Carol Calin · Data Scientist & AI BuilderDate  June 2026 · Bucharest ⟷ Amsterdam
The situation

A niche bank with big ambition — AI as a multiplier, not a replacement for your specialists.

BRCI's edge is the relationship: “Ask a specialist”, advisory for SMEs, trust built client by client. The right AI strengthens exactly that relationship — it gives advisors faster, safer answers without taking the human out of the loop. The starting point isn't a model; it's a team that knows how to use one.

I.

Two branches, the rest digital

With branches in Bucharest and Galați, your growth comes from digital channels: online business accounts, iBanking, Mobile, instant payments. This is exactly where enablement separates a tool you bought from a team that uses it every day.

II.

Prudent budget, small steps

At the scale of a niche bank, transformation isn't a big, risky project. It's a small, measurable pilot on a single workflow — value proven before any rollout.

III.

Fraud isn't just a “fake face” anymore

You already send clients alerts about fraud and social engineering. Attackers now use camera injection, face swapping and genuine IDs with a synthetic face — exactly what a digital onboarding must catch, and what I studied at the NFI.

From people to systems

Four workstreams, in the order that matters for a relationship bank: people first, then systems.

I lead with enablement — a tool the team doesn't use is money wasted. Then we build gradually, only what has proven itself on a real workflow.

01 — Enablement · people first

“Ask a specialist”, backed by AI — no hallucinations in front of the client

Hands-on sessions with your advisors: how to ask an agent for a draft answer grounded in internal procedures, how to check the source and citation before giving it to an SME client, and what should never be sent into a public tool. I don't teach engineers; I teach people — and together we build reusable “skills” for your real tasks.

Why it fits BRCIYour brand is the relationship — “grow together”. Enablement keeps the human at the centre and makes them faster, instead of replacing them.
What I build onA psychology and community background — hands-on teaching for people, not engineer-only slide decks.
02 — Agents for onboarding & digital support

iBanking, Mobile, online business accounts — more throughput, no extra headcount

Agents that prepare draft replies and triage requests across iBanking, Mobile and online SME onboarding — the human approves, the agent types. With only two branches, this absorbs volume without opening new counters.

Why it fits BRCIYour growth is digital. Less repetitive manual work, faster answers for business clients.
What I build onMulti-agent knowledge bases with semantic indexing, citations and access control over private documents.
03 — KYC & anti-fraud · forensic-grade

AI fraud your standard check won't see — from camera injection to real ID + synthetic face

Not a generic “fraud model”. Defence layers tuned to the mechanics I studied at the NFI: video-manipulation detection, inconsistencies between the presented document and the live stream, signals of software injection in place of a physical camera, and alerts for synthetic documents that pass OCR but fail forensic analysis.

Why it fits BRCIYou already issue fraud and social-engineering alerts to clients. A check that was “good enough” two years ago can become the door to a regulatory incident.
What I build onResearch on video manipulation and forged-document detection at the NFI; thesis defended at the Dutch Ministry of Justice & Security.
04 — Build · optional, after enablement

A sprint on a chosen workflow: reporting, instant payments or back-office

When a workflow is worth automating end-to-end, I deliver it as a solo builder with team throughput, using coding agents (Claude Code, Cursor). Small, bounded, measurable — on infrastructure you control, not a monolith project.

Why it fits BRCIPrudent budget: delivery in pieces, each with a visible result, started only after enablement has shown where the value is.
What I build onEnd-to-end solo delivery with coding agents; finalist at 10+ international hackathons.
Realistic about scale and budget

A small bank doesn't need a big AI budget. It needs good choices.

Many firms either avoid AI or burn money sending every task to the most expensive cloud model. At your scale, neither makes sense. I work the other way: start small, measure, and let the result justify the next step.

People first. A tool the team adopts is worth more than ten that go unused. Then the right automation. Repetitive tasks run on models hosted at your side; client data never leaves the bank. And one measure that matters: time saved and risk reduced, stated in management's language, not in jargon.

How effort is allocated
Drafts, summaries, KYC pre-checksLocal, on your hardware
Any client dataStays with you, on-prem
Rare, complex reasoningFrontier model, anonymized
The first engagementSmall pilot, clear criteria
The report to managementTime saved, no jargon
Why me

I teach people, I build systems, and I've worked on fraud where the evidence has to hold up.

I

Enablement that sticks

I teach from people, not from code — a psychology and community background. Adoption lasts because the team understands, not just installs.

II

Forensic-grade fraud

Research on video manipulation and forged-document detection at the Netherlands Forensics Institute (NFI); thesis defended at the Dutch Ministry of Justice & Security. I know identity fraud by concrete vectors, not as a slogan.

III

Privacy-preserving by design

ML on sensitive data with homomorphic encryption (PySyft, TenSEAL) and zero-knowledge ML (EZKL). Models work without exposing the data underneath — exactly a bank's constraint.

IV

One builder, team-scale output

End-to-end solo delivery with coding agents (Claude Code, Cursor); finalist at 10+ international hackathons. Speed without a large team is the whole point at your scale.

Plain and honest: I have not worked inside a bank, and I won't pretend otherwise. What I bring is the part that's genuinely hard to hire — enablement that sticks, forensic-grade fraud detection, and the speed to ship — and I'll learn the banking specifics from your team. The domain I'll learn from you; the technology I already know cold.
A low-risk way in

No big-bang. Start with people, prove value on one workflow, scale only what earns it.

01
Step one · ½ day · complimentary

Enablement workshop + workflow map

A short session with the “Ask a specialist” team and back-office: I show concretely how to use AI safely and we map where it helps at BRCI and where confidentiality requires it to stay local. You leave with a written plan — whether or not we continue.

02
Step two · ~2–3 weeks · fixed fee

One workflow, fully working, fully private

I build the first pilot — most likely the procedures copilot with citations for advisors — on infrastructure you control, with success criteria agreed up front. Small, bounded, and yours.

03
Step three · ongoing · only if it earns it

Scale across onboarding, KYC and reporting

We roll the proven pattern further; train the rest of the team; add the anti-fraud layers. Retainer or per-project — your call, reviewed periodically.

Letter of intent
BRCI
Bucharest · Șos. Ștefan cel Mare nr. 3
June 2026
Confidential

For the attention of the executive management of Banca Română de Credite și Investiții (BRCI).

Dear Sir or Madam,

I've followed BRCI as a niche bank relaunched in 2014, with roots going back to 1990, that openly commits to a promise rare in the market: “We want to grow together.” You grow without a large branch network, betting on relationships and digital channels — exactly the context where the right AI helps most.

I propose we start with people: training the “Ask a specialist” team to use AI with confidence, then adding agents for digital onboarding, for compliance triage, and an anti-fraud defence tuned to real identity attacks — the field I researched at the Netherlands Forensics Institute.

This isn't a pitch sent in bulk: I wrote it for BRCI, in your colours, starting from your products and priorities. I propose a small, no-risk first step — a half-day of enablement and a workflow map — so you can see the value before any commitment.

I remain at your disposal, in Romanian or English, for a short conversation whenever it suits you.

Carol Calin
Data Scientist · AI, privacy-preserving ML & enablement · Amsterdam
Get in touch

A 30-minute conversation — you'll leave with ideas, not a form.

No obligation. I'll come with two or three concrete ideas drawn from BRCI's own products: “Ask a specialist”, digital onboarding, and anti-fraud defence.