Forward Deployed Engineer · Applied AI
Hi, I'm Federico.Physicist by training, engineer by trade.
I design, deploy, and scale agentic AI systems for enterprise customers — from technical discovery and architecture through evaluation and production. Currently a Forward Deployed Engineer at Wonderful AI. Physicist by training, with earlier work in HPC and polymer dynamics.
Production voice agent designed and scaled for a major telecom enterprise.
About me
I studied physics — modeling complex systems, running large-scale simulations, and learning to reason from first principles. Today I apply that mindset as a Forward Deployed Engineer at Wonderful AI, where I'm embedded with enterprise customers to scope, architect, deploy, and evaluate production AI systems. I've worked across customer-facing voice agents and back-office automation, including scaling a telecom voice agent to 100,000+ calls per day. Previously I built enterprise data products at Deloitte and founded IncognitoAI, a privacy layer for generative AI. I like working at the boundary between customers and engineering: understanding a real operational problem, designing the system, and getting it to work reliably in production. Most of my recent code lives in private enterprise repos under NDA — for context on style and approach, see the projects and published physics paper below.
Skills
Applied AI
Backend & APIs
Data & Infrastructure
Scientific Computing
Projects
Enterprise Voice Agent
Production agentic voice system deployed for a major telecom. Scaled to 100K+ inbound calls/day with 75% containment and 1.3s median latency. I owned the deployment end-to-end: solution design, tool/API integration with enterprise systems, evaluation harness, monitoring, and CI/CD.
arte — Typewriter Literary Rendering
An image and a full book go in; a piece of physical art comes out. The tool maps a novel's characters onto an image grid — Floyd–Steinberg dithering, adaptive thresholds matching the image histogram to the text's character-weight distribution — consuming the text strictly sequentially, never reordering a single letter. Zoom in and it's readable Melville. It then emits vector stencil PDFs at true physical scale, auto-tiled across A2/A1/A0 sheets for laser cutting.
Rendered from Moby-Dick (1851) over a public-domain humpback photograph — NOAA / National Marine Sanctuaries.
IncognitoAI
A privacy proxy for LLMs that detects and anonymizes PII before inference. Uses a fine-tuned Llama 3B for entity recognition, with regex pre-filters for structured patterns. Sensitive tokens are replaced before reaching third-party LLM providers and re-injected in the response.
Real Estate Valuator
Internal Deloitte tool for automated property appraisal, combining OMI comparables, cadastral GIS data, and statistical regressions (monoparametric + power law). Predictive analytics meets territorial intelligence.
Experience
Forward Deployed Engineer, Wonderful AI
Embedded with enterprise clients to scope agentic AI use cases and own deployments end-to-end across solution design, integrations, evaluation, monitoring, and CI/CD. For a major telecom, designed and scaled a customer-facing voice agent to 100,000+ calls/day in three months, achieving 75% containment and 1.3s median latency.
Data Scientist, Deloitte Business Solutions
Built an automated real-estate valuation engine, managing backend logic, pricing algorithms, and ETL pipelines for geo-spatial data. Developed Python + PostgreSQL pipelines integrating enterprise data from 30+ plants worldwide to power live C-level dashboards.
Founder, IncognitoAI
Built a proxy system that anonymizes PII in GenAI prompts using open-source models. Led product vision and early sales. Won July Demo Day, secured €20k in public grants.
Researcher, CINECA (ISCRA Class B)
Engineered high-performance computing workflows and ran large-scale simulations in polymer physics. Optimized data pipelines using the Leonardo HPC framework.
View publication →
Get in touch
Have a question, a project idea, or just want to connect? Drop me a message.