Technical & Strategic AI Consulting
Assessment of your current business processes, identification of automation opportunities, and architectural design of AI-driven strategies.
Rate: 30 € / hour
Bridging the gap between advanced AI research
and practical business automation.
With a foundation in Applied Mathematics and Computer Science (St. Petersburg State University) and a recent Master's degree with honors in Artificial Intelligence and Cybersecurity Engineering from Universitat Rovira i Virgili (Tarragona), I bring a robust technical skillset to the table. My expertise encompasses LLMs, Python, PyTorch, TensorFlow, LangChain, RAG, and multi-agent systems like CrewAI.
What sets me apart is my 10 years of prior experience in management and operations across the B2B and B2C retail and insurance sectors. I deeply understand business processes and data analysis, which allows me to implement AI not just for the sake of technology, but to solve real business problems. Having successfully completed a practicum implementing Computer Vision on a manufacturing line at IPTE Spain (Reus), I am fully equipped to design, build, and deploy intelligent automation systems from the ground up.
Services & Offerings
I provide 100% remote B2B consulting and implementation services aimed at optimizing and automating your business workflows using cutting-edge AI technologies.
AI & LLM Implementation
Integration of Large Language Models, Retrieval-Augmented Generation (RAG), and multi-agent systems into your current workflows for text, data, and communication automation.
Turnkey projects: 50 € – 1,500 €
Computer Vision Solutions
Full-cycle training and deployment of Computer Vision models for manufacturing, quality control, object detection, and visual data extraction.
Turnkey projects: 100 € – 2,000 €
Portfolio
Selected projects I have delivered, from a production-line deployment to multi-agent LLM systems. Click an image to see more.
4
Automated Weld Verification on a Production Line
A two-stage deep learning pipeline built for IPTE Spain (Reus): an RT-DETR detector finds each weld, then a U-Net segments it pixel by pixel and produces a heatmap the operator can read. It reaches a Dice score of 0.92 in real time, and a human-in-the-loop workflow adapts it to a new weld type with about 50 labelled images.
3
Counting Parts in Thermal Images
A classical computer vision system that counts complete parts on a heat-sealing line from thermographic camera images. Background flattening, Hough line and circle detection and edge filtering keep only fully visible parts; the parameters were tuned automatically and the count was correct on all 7 sample images.
Multi-Agent Article Writing with CrewAI
A team of three LLM agents (researcher, writer and editor) turns a topic into a publication-ready article. The researcher drafts an outline that a human approves, the writer produces a draft of 800+ words and the editor checks clarity, accuracy and grammar. Shared memory runs on local Ollama embeddings.
3
Crack Detection on Concrete Surfaces
A deep learning model that finds cracks pixel by pixel in photos of concrete and asphalt, for automated inspection of buildings and infrastructure. It is based on the DeepCrack architecture (a VGG-16 network with multi-scale side outputs), trained in PyTorch on 300 annotated images; on 237 unseen test images it reaches an F1 score of 0.71, even though cracks are only a few pixels wide.
2
Multi-Agent Mission Planning
Architecture for a CrewAI system that plans a Mars exploration mission for rovers, drones and satellites. Six crews of specialised agents share the work: one prepares the mission data, four plan the routes for each vehicle type, and one merges everything into a single plan that respects energy, time and terrain constraints. Designed in a team of five.
2
Statistics on Encrypted Data
A web app that lets an untrusted server calculate statistics on sensitive data without ever decrypting it. It implements the Domingo-Ferrer privacy homomorphism in Python (200-digit keys, probabilistic encryption) and computes means, covariance and a regression on an encrypted medical dataset; only the data owner can read the results. Developed in a team of three.
Get In Touch
Open for new projects and B2B contracts. Whether you are looking for strategic advice on adopting AI or need hands-on development for a specific automation or computer vision task, I am ready to help your business grow efficiently.