DAVIDE MODOLO

AI & Machine Learning Engineer, R&D

OpenCity Labs (Startup) | Remote | August 2025 - Present

• Single-handedly engineered an NLP pipeline for RAG ingestion with PII anonymization via a fine-tuned Transformer-based NER model and an in-memory reversible dictionary, processing 50k+ daily inputs.
• Built a spaCy detection + LLM fallback guard to fix a cross-lingual model failure in a RAG pipeline: despite explicit English instructions, Italian-only retrieved context caused the model to default to Italian in 35-40% of non-Italian queries, eliminating the error. Released to 10+ clients.
• Developed an open-source agentic AI system. Designed hybrid RAG architecture (vector + full-text reciprocal rank fusion) to solve domain-specific retrieval failures: pure embedding search missed legal acronyms and regulatory references, while keyword search failed on semantic ones. Built MCP-based tool-calling infrastructure with human-in-the-loop approval, reducing targeted workflows from days to hours.
• Dockerized all services with GitLab CI pipelines automating testing and Docker image builds for consistent, reliable deployments.

Intern / AI Engineer

Eurecat Technology Center | Barcelona, Spain | April 2024 - June 2024

• Implemented a token-level uncertainty estimation framework for planning tasks by analyzing log-probabilities, benchmarking performance across proprietary (GPT-4o) and locally quantized models (Llama 3.1 via llama.cpp).
• Developed a RAG system utilizing gte-large embeddings to retrieve context-aware user preferences, significantly reducing model hallucination in recurrent decision-making scenarios.
• Prototyped a multimodal agent integrating LLaVA for visual reasoning tasks and validated the uncertainty reconstruction approach across multiple simulated environments.

Master's Degree in Artificial Intelligence Systems

University of Trento | Trento, Italy | September 2021 - March 2025

• Thesis: Exploring the Use of LLMs for Agent Planning: Strengths and Weaknesses - investigated how large language models perform on planning and decision-making tasks, identifying capabilities and failure modes. Developed a rigorous uncertainty quantification methodology based on token log-probabilities and conformal prediction, generating heatmap visualizations across 36 test cases (3 LLM versions, 4 map sizes, 2 tasks) to characterize spatial reasoning quality across the entire state space.
• Relevant Courses: Machine Learning, Deep Learning, Natural Language Understanding, Automated Planning, Law & Ethics in AI, Fundamentals of AI, Signal Processing, AI for Finance, High Performance Computing.

Bachelor's Degree in Computer Science

University of Trento | Trento, Italy | September 2017 - June 2021

• Thesis: Healthy Plus - Redesign and evolution of an Android application for monitoring healthy lifestyles.
• Core Topics: Algorithms & Data Structures, Software Engineering, Database Systems, Operating Systems, Computer Architectures, Distributed Networks.

• Built a multi-stage deep learning pipeline to classify Lung Ultrasound frames into 4 severity scores (0-3) using a fine-tuned ResNet18 on ~47k frames from 11 patients, with patient-based train/test splitting to prevent data leakage.
• Added a binary confidence classifier that detects uncertain predictions by analyzing softmax output patterns, and a t-SNE-based image similarity retrieval module that finds visually similar training images when the primary classifier is not confident.
• Designed an aggressive augmentation pipeline (affine transforms, brightness scaling, Gaussian blur, horizontal flips) to combat overfitting on the limited medical dataset.

• Implemented 5 neural architectures from scratch in PyTorch for multi-task joint intent classification and slot filling on ATIS and SNIPS datasets: baseline LSTM, bidirectional LSTM, encoder-decoder, and fine-tuned BERT and ERNIE 2.0.
• Custom encoder-decoder achieved the best overall results (94.95% slot F1, 96.42% intent accuracy on ATIS), outperforming pretrained transformers which suffered from a vocabulary mismatch when custom slot-label tokens were added to their tokenizer.
• Each model trained 5 times with different random seeds and reported with mean ± standard deviation; used early stopping and custom LR scheduling (halve every 3 epochs without improvement).

• Built a full unsupervised domain adaptation pipeline in PyTorch based on the HoMM paper, matching 3rd and 4th-order statistical moments between source and target feature distributions to capture shape differences (skewness and kurtosis) that simpler methods like MMD and CORAL miss.
• Used ResNet34 pretrained on ImageNet with a modified tanh activation layer, balancing three loss terms: cross-entropy (source supervision), HoMM discrepancy (domain alignment), and discriminative clustering with moving-average centroids (target pseudo-labels).
• Achieved +11.53% gain over the source-only baseline on the Adaptiope dataset (20 object classes across product_images and real_life domains).

• Developed an autonomous software agent in Node.js for a grid-based parcel delivery game using a Belief-Desire-Intention architecture with intention revision and plan caching.
• Integrated a PDDL online solver (planning.domains API) for declarative path planning, replacing an earlier A* approach that required hand-coded edge cases for each map configuration.
• Built two multi-agent coordination strategies: angular map slicing with basic communication, and intention sharing with conflict resolution via mutual scoring.

• Implemented parallel solutions to the closest pair of points problem in up to 5 dimensions using C and MPI, with both brute-force (O(n²)) and divide-and-conquer (O(n log n)) algorithms under a master-worker model.
• The brute-force approach scaled near-linearly to ~28× speedup on 80 cores (embarrassingly parallel), while divide-and-conquer plateaued at ~7.3× due to cross-boundary strip-exchange communication overhead, a tradeoff between asymptotic complexity and parallel scalability.
• Handled the INT_MAX MPI buffer limit with automatic multi-tranche serialization of variable-size Point structs via MPI_Pack/MPI_Unpack; benchmarked up to 250 million points across 1-80 cores on an HPC cluster.