The Bench

A core team of operators. Assembled around your problem.

Below is some of our core team. Behind it sits a wider network of senior consultants we plug in by specialism — so every engagement is staffed by the people whose specific experience makes the difference, drawn from Microsoft, X& Financial, Oxford, Cambridge, LSE, and the Serbian AI Institute.

LB

Cambridge · Microsoft AI · Sanbo / X& Financial

Luka Bojović

Founder · Project Lead & AI Architecture

Luka leads engagements end-to-end. He read engineering at Cambridge, spent years at Microsoft as a Senior Program Manager shaping the AI experiences inside Windows and Copilot+ PC, and most recently led product on Sanbo — an AI-native platform for private capital markets. He is the architect on most Aries engagements: framing the business case, designing the AI module, and owning the bar for evaluation and audit.

LinkedIn ↗

Expertise

  • ·AI product strategy
  • ·Document & question-answering workflows
  • ·Eval frameworks
  • ·Enterprise delivery
SB

Office of the Prime Minister · Hub 201

Stefan Badža

AI Strategy · Public Sector & Enterprise Security

Stefan brings the security and governance lens that lets AI ship inside regulated environments. He previously served in the Office of the Prime Minister of Serbia and helped build Hub 201, one of the region's leading tech and innovation hubs. He partners with public-sector and enterprise clients on strategy, secure architecture, identity and access, and the controls that make a model usable in production.

LinkedIn ↗

Expertise

  • ·AI strategy & roadmaps
  • ·Public sector engagements
  • ·Enterprise security & governance
  • ·Secure deployment
MV

Telesign · X& Financial

Marijana Vujadinović

Data Science & Machine Learning

Marijana is our data scientist and ML engineer for production-grade modelling problems. At Telesign she worked on fraud and risk signals at global messaging scale, and at X& Financial she built data science and ML systems for private capital markets. She now leads our applied ML work — from feature engineering and model selection through evaluation and deployment.

LinkedIn ↗

Expertise

  • ·Applied machine learning
  • ·Fraud & risk modelling
  • ·Feature engineering
  • ·Model evaluation & deployment
AM

Enterprise BI · Modern data stack

Andrea Miljuš

Data Warehouse & Business Intelligence

Andrea owns the layer that everything else stands on. She designs and builds the warehouses, semantic models and BI surfaces that turn fragmented operational data into something an AI module — or an executive — can actually trust. Without her, copilots hallucinate; with her, they cite.

LinkedIn ↗

Expertise

  • ·Data warehousing
  • ·Modern data stack & ELT
  • ·Semantic models
  • ·BI & executive reporting
UR

Oxford · Microsoft Development Center

Uroš Ristivojević

Data Science · Proprietary LLM Training

Uroš is our specialist for clients who want — or need — a model of their own. He has trained, fine-tuned and evaluated LLMs at production scale, and brings the rigor of a research engineer to the question of when a custom model is genuinely worth it versus when an off-the-shelf one will do.

LinkedIn ↗

Expertise

  • ·LLM training & fine-tuning
  • ·Model evaluation
  • ·Scalable architecture
  • ·Applied research
MB

LSE · Deloitte · FIS · Hyperoptic

Maja Bojović

Business Strategy · Analytics & Finance

Maja makes sure every engagement has a number attached to it. She joins from senior strategy, analytics and finance roles across consulting and high-growth telecoms, and runs the commercial framing of our work — the business case, the unit economics, and the metrics that decide whether a pilot becomes a programme.

LinkedIn ↗

Expertise

  • ·Business strategy
  • ·Commercial analytics
  • ·Financial modelling
  • ·AI ROI & business cases
PC

Manufacturing · Quality control systems

Predrag Cvetković

Industrial Computer Vision

Predrag brings the camera into the factory. He has shipped industrial computer-vision systems for quality control, defect detection and process monitoring across heavy manufacturing — the kind of work where a model has to be right at line speed, on a real shop floor, every shift.

LinkedIn ↗

Expertise

  • ·Industrial computer vision
  • ·Quality control automation
  • ·Edge deployment
  • ·Manufacturing analytics

Working together

Tell us the problem. We'll bring the right team.