AI assistants and agents
Conversational and multi-agent workflows that synthesize information, support decisions and make domain analytics accessible through natural language.
- LLM assistants
- Multi-agent workflows
- Text-to-SQL
- Tool calling
Lead AI EngineerApplied AI Consultant
Cut complexity.Deliver AI products that perform.
Hi, I'm Vanja. I help product and engineering leaders turn ambiguous AI opportunities into shipped capabilities — from executive assistants and document automation to analytics agents, anomaly detection and risk models. I bring hands-on engineering, product development and technical leadership shaped by 15+ years across startups, scale-ups and enterprise teams.
15+years across data, ML and AI product delivery
~6 weeksfrom early requirements to an active AI-assistant rollout
24xreduction in monthly manual invoice-extraction work
5M+credit decisions supported each year across 13 markets
Services
From unclear opportunity to useful product. I work with product and engineering teams where AI needs to move beyond a demo — into a system that is grounded, evaluated, maintainable and valuable to the people using it.
Document and knowledge automation
Reliable pipelines that turn PDFs and business documents into validated structured data or grounded answers.
- Document extraction
- Structured outputs
- OCR and multimodal AI
- RAG
Applied machine-learning products
Customer-facing and operational ML capabilities for monitoring, risk, classification, prediction and prioritization.
- Anomaly detection
- Risk modeling
- NLP classification
- Regression
AI product discovery and delivery
Pragmatic support from ambiguous opportunity to evaluated prototype, technical plan, MVP and production capability.
- Product discovery
- Evaluation design
- Rapid prototyping
- Technical roadmap
Skills
A practical stack for shipping applied AI. Technologies are selected around the product and its constraints. These are the areas most relevant to the work I deliver — not an exhaustive tool inventory.
AI product systems
- LLM applications
- Multi-agent workflows
- Structured outputs
- RAG and embeddings
- Text-to-SQL
- LLM evaluation
Product and backend delivery
- Python
- FastAPI
- Pydantic
- Docker
- GCP Cloud Run
- Testing and CI
Applied machine learning
- scikit-learn
- XGBoost
- Meta Prophet
- Anomaly detection
- NLP
- Model explainability
Data engineering and analytics
- SQL
- BigQuery
- Redshift
- Airflow
- PySpark
- ETL pipelines
Case Studies
Seven products, each grounded in a real operating problem.
CEO-facing AI Executive Assistant
AI Invoice PDF Parser
Text-to-SQL AI Payments Analytics Agent
Payments Anomaly Detection Product
Credit Default Risk Modeling
Manufacturing Cost-Savings Regression Model
Manufacturing Sales-Prioritization Map
Details
Experience, leadership and background.
Selected Experience
Product work across different operating realities. Startup urgency, scale-up growth and enterprise constraints require different choices. My work spans each — including two successful acquisition journeys from Instantor to Tink and Tink to Visa.
EQTAI Engineering Consultant via Northbridge
PagosLead Data Scientist & Engineering Manager
TinkSuccessfully ExitedSenior Data Scientist
InstantorSuccessfully ExitedSenior Data Scientist
SandvikData Scientist
EricssonResearch Engineer & PhD Candidate
Leadership
Close to the code. Clear on the outcome. My role often sits across engineering, product and commercial context: shaping what to build, making the technical choices and helping the team deliver it well.
Technical and product leadership
- Led applied AI, machine-learning and data-science delivery from discovery to shipped products.
- Turned ambiguous business problems into scoped technical plans, prototypes and production capabilities.
- Balanced technical quality, delivery speed and business value across startup, scale-up and enterprise settings.
- Owned technical roadmaps and cross-functional delivery for B2B AI, ML and analytics products.
People and stakeholder leadership
- More than two years of engineering-management experience, leading a team of up to four.
- Hired, onboarded, coached and managed senior, mid-level and junior data scientists.
- Worked directly with founders, customers and senior leaders including CEO, CTO, Chief of Staff, CPO and VP-level stakeholders.
- Built ways of working that strengthened execution, ownership and individual growth.
Foundation
Research depth with a product bias.
PhD in social network analysis
University of Zagreb, focused on graph algorithms, statistical validation and data-processing methods for measuring user influence.
Eight peer-reviewed papers
Academic work included an article in Complexity, alongside six years combining research engineering with doctoral work.
Recognized for execution and ideas
Honorary Doctoral Commencement Speaker, Deezer Music Hackathon winner and Ericsson Global Innovation competition winner.
Contact
Have an AI product problem worth solving?
If you are evaluating an applied AI opportunity, unblocking a product, or looking for hands-on technical leadership, email is the simplest place to start.
vanja.smailovic@gmail.com