LLM & NLP Integrations
Custom integrations with OpenAI, Anthropic, open-source models. Prompt engineering, function calling, structured outputs and evals.
Freelance · P.IVA
I help companies and startups ship production-grade AI features — NLP pipelines, RAG systems, LLM integrations and the backend that makes them work in the real world.
Services
Custom integrations with OpenAI, Anthropic, open-source models. Prompt engineering, function calling, structured outputs and evals.
End-to-end Retrieval-Augmented Generation pipelines — ingestion, chunking, embeddings, vector DBs and answer evaluation.
Production-ready Python and Java backends. FastAPI services, PostgreSQL, Docker deployments — built lean, no over-engineering.
Workflow automation for SMEs combining n8n, LLMs and external APIs. Replace repetitive ops work with reliable pipelines.
Custom models for classification, time-series and computer vision. From dataset cleaning to deployment, with proper evaluation.
Code reviews, architecture audits, AI-readiness assessments. Honest feedback on what to build, what to buy and what to drop.
Selected work
A selection of recent projects across data engineering, computer vision and deep learning.
End-to-end data cleaning, quality assessment and enrichment pipeline on a real-world messy dataset. Fuzzy record linkage, schema validation and quality scoring across 6,000+ records.
Breast cancer tissue classification using fine-tuned EfficientNet, ResNet and Vision Transformer (ViT). Applied Grad-CAM for model interpretability.
BiLSTM / Conv1D / GRU ensemble with attention mechanisms for sequence modelling. Achieved 95.5% F1 on the AN2DL competition.
More on github.com/devisnsk
About
I'm a Computer Science master's student at Politecnico di Milano, specialising in AI/ML, NLP and Edge AI. Alongside my studies I work as a freelance AI & Software Engineer, helping companies turn AI ideas into production systems that actually ship.
Get in touch
Reach out for freelance work, collaborations, or just to say hi. I usually reply within 24 hours.