MapTiler
Geospatial
Geocoding & maps search at planet scale
Develop structured search, routing and address-intelligence features on a geocoding service indexing billions of geographic features - improving deduplication and postal-code validation, fixing production bugs, and scaling load-tested infrastructure as traffic grew.
Structured search
Routing
Python
PostGIS
Mesh Solutions
Geospatial
Address parsing & normalization with libpostal
Designed and built services around libpostal to parse, normalize and standardize messy international addresses into structured components (street, city, postcode) - powering accurate geocoding, deduplication and data-quality workflows for logistics, CRM and KYC use cases.
libpostal
NLP
Geocoding
Python
Client project
Geospatial
Address dataset cleanup for geocoding
Cleaned and standardized a messy client address dataset so it could be geocoded accurately - improving match rates and downstream delivery and analytics.
Geocoding
Data quality
Python
MapTiler
Data & ML
Release process & geocoding quality
Set up the release process for geocoding and index updates. Built automated regression and data-quality tooling that guards against map vandalism and catches regressions before major releases ship.
Release engineering
QA automation
TypeScript
Client Project
Data & ML
Customer segmentation analytics
Built data models and analytical insights that turned raw usage data into clear customer segments - enabling data-driven product decisions across the team.
Analytics
Python
SQL
Open source
Data & ML
Anomaly Detection Platform
Generate, analyze and benchmark different anomaly-detection algorithms on transaction data - a hands-on learning and comparison platform.
Python
ML
GitHub →
Open source
Data & ML
BERT fine-tuning
Training and evaluating a BERT-based model for multilabel text classification in TensorFlow, predicting multiple labels per input.
NLP
TensorFlow
BERT
GitHub →
Open source
Data & ML
Social media post classification
SVM/XGB classifiers exposed through a FastAPI endpoint, dockerized and deployed - a compact end-to-end ML service.
FastAPI
Docker
ML
GitHub →
Personal
Data & ML
Bosnian News Analytics
A scraping pipeline, analytics dashboard and sentiment analysis across multiple news portals, each with its own profile and statistics.
Django
PostgreSQL
NLP
Open source
Data & ML
Random String Detector
An open-source Python package that flags random or gibberish strings - handy for validating and cleaning text and address data before processing.
Python
PyPI package
Data quality
GitHub →
Client project
Data & ML
Mobility data analysis
Data analysis and processing for a mobility company - turning large movement and operational datasets into clear, decision-ready insight.
Python
BigQuery
Analytics
Open source
Data & ML
Large Dataset Filtering
A desktop tool that rapidly queries and filters multi-GB CSV/Excel datasets and exports results - fast where web apps struggle.
Python
Tkinter
GitHub →
Venture · Founder
Product
DeliveryMesh
An optimization and delivery-planning platform that helps logistics teams cut driving time and cost. Routing engine, spatial data model and planning UX - built from idea to product.
Route optimization
Full-stack
deliverymesh.com →
Venture · Founder
Product
Mesh Solutions
My independent software & data practice - custom applications, geospatial features (geocoding & routing) and clean, standardized address/POI datasets for clients across the US and EU.
Custom software
Geospatial
Data
meshdatasolutions.com →
Productized service
Product
Geo-Data Quality Audit
A packaged engagement that audits a client's address/POI data and delivers a concrete action plan to raise geocoding accuracy and data quality.
Geospatial
Data quality
Consulting
Co-author · Research
Research
PhD pharmacogenetics research
Long-running support on a PhD study of genetic biomarkers behind drug-treatment response. I built a web tool for working with the Case Report Form data (insert, preview, edit, analyze, with anomaly detection) and support the statistical analysis and AI-assisted advanced analysis.
Python
Django
Statistics
AI / ML
University of Sarajevo
Research
UAV 2D/3D semantic segmentation
Master's research on 2D/3D image analysis from UAV imagery - segmentation, geocoding and change detection - including a novel colour-based segmentation method modeling lighting with neural networks.
Computer Vision
OpenCV
3D Segmentation
Atlantbh
Research
Anomaly detection for manufacturing
Applied anomaly-detection techniques to semiconductor manufacturing data to surface defects and irregularities early in the process.
Python
Anomaly detection
University of Sarajevo
Research
Chronic Kidney Disease - ML
Finding the significant attributes for CKD with statistics and machine learning, helping clinicians act preventively to preserve basic kidney function.
Python
Pandas
ML
Co-mentor · Univ. of Zenica
Research
RAG for technical documentation search
Long-time mentor on a bachelor's thesis building a Retrieval-Augmented Generation system that indexes technical documentation as vector embeddings for semantic search - steering the direction toward RAG, vector databases and applied AI.
RAG
Vector databases
LLMs
Co-author · Research
Research
Deep learning for breast cancer staging
Co-authoring a narrative review of deep-learning models that predict axillary lymph node metastasis from breast histopathology slides - leading the software engineering and AI/ML side of the work alongside a medical collaborator.
Deep Learning
Medical imaging
AI
Co-mentor · Univ. of Zenica
Research
ML address parsing for Bosnia & Herzegovina
De-facto mentor on a master's thesis developing and evaluating machine-learning models for parsing Bosnian addresses - including the first representative BiH address dataset for geocoding and GIS. Helped define the topic and guide the project throughout.
NLP
Address parsing
Geocoding
