manuel scionti · ai engineering · forward deployment

AI systems, from ambiguous problem to production.

I build the part of AI systems that comes after the demo: production RAG, agentic systems, and the evals that keep them honest.

I partner with stakeholders from discovery and solution design through implementation, rollout, and adoption - while staying hands-on with the code.

now: Senior Data Consultant at Capitole, focused on Applied AI and client delivery. My most recent engagement was with ADP’s Core AI Platform team, ending in July 2026 after 17 months.
how I work: Discovery → feasibility → architecture → implementation → rollout → post-sales. I am most useful where the customer context and the engineering work cannot be separated.
before: Customer-facing AI delivery at Neodata: joined sales in first client meetings, led technical discovery, and owned delivery and relationships across 20+ enterprise opportunities. ML infrastructure on AWS at Koexai.
community: Community manager of Applied AI Engineers ↗, helping software and data engineers move into production AI and forward-deployed roles.
writing: One Token at a Time ↗ — essays on the engineering and product decisions behind production AI.
open to: Applied AI, Forward-Deployed Engineering, and Customer Engineering roles across Europe. EU citizen; open to relocation.

selected work

Enterprise AI platform delivery at ADP — shared production capabilities across 10+ business units and two regions.
Catania Airport media intelligence — end-to-end ownership with monthly CFO and CTO roadmap reviews.
Forward-deployed AI at Neodata — from the first sales conversation through implementation, rollout, and post-sales.
Lucia, a bilingual screening agent — deterministic flow; the LLM’s only job is phrasing.

Conversational AI is a UX problem ↗ — why control flow, ambiguity handling, and the first message can matter more than another model upgrade.