Agentic AI · Energy systems · Technical leadership
I’m Giuseppe, an AI & Tech Solutions Lead. I lead a team of five building agentic systems for energy operators across the US, UK and Ireland. From typed tools and MCP to evaluation and human approval, I connect production AI with the people and decisions it serves.
“Why is Site B underperforming since Tuesday?”
Step budget7 / 8
Token budget46%
resolve_outcomeplangoal="explain underperformance" site=B window=7dScoped to one site and one week, energy impact first.assess_gapsplanhave=[alerts] need=[events, power_curve, curtailment]Three evidence gaps to close before any answer.events.querytyped toolsite="B", window="7d", rank_by="energy_exposure"Open events ranked by lost energy.telemetry.comparetyped toolassets=["T-07","T-11"], signal="power_curve"Two turbines below their expected power curve.authorizeguardrailscope="read" mutation=falseRead-only access confirmed. Nothing changed.reassessplangaps_closed=3/3 budget_left=1Supported hypothesis: yaw misalignment on T-07 and T-11.propose_changeneeds approvalaction="create_work_order" assets=["T-07","T-11"]Inspection work order drafted. Waiting for a person.Needs your confirmationCreate an inspection work order for T-07 and T-11
Question → Evidence gaps → Typed tools → Human decision
Conceptual illustration of a WaveAI run · No client data · Decisions stay with people
per site investigation with WaveAI, down from about an hour
to onboard a customer with the MCP-first workflow, down from about three weeks
assets per customer ranked daily, with triage in minutes, down from about an hour
Also on the record
01 / Selected work
Prioritize what matters. Investigate with evidence. Deliver through reviewable tooling. Choose a product below to trace its part in the loop, while authority and consequential decisions stay with people.
A bounded agentic copilot that investigates operational questions with evidence, explicit authority, and replayable runs.
Investigations ~1 hr → minutes · 2× weekly users
Read case studyAn explainable daily agenda for teams operating fleets of 100K+ assets per customer.
Triage ~1 hr → minutes · 4 teams
Read case studyA reusable, reviewable workflow connecting client specifications, analytics logic, and supported delivery operations.
Onboarding ~3 wks → 3–4 days
Read case studyAgentic copilot for energy operations
A bounded investigation loop on LangGraph and MCP: resolve the outcome, assess evidence gaps, select a typed tool, validate and authorize, then reassess within explicit step and token budgets.
What I built
Outcome
~1 hr → minutes
Per site investigation. Weekly active users doubled within a quarter.
Read the WaveAI case studyPrioritization at fleet scale
Classifies the full open event backlog every day, tracks age and energy exposure, and ranks sites separately for Operations, Performance Engineering, Asset Management and Executives.
What I built
Outcome
~1 hr → minutes
Triage time, deployed across the customer base.
Read the Smart Priority Tracker case studyReviewable delivery tooling
An MCP server and skill exposing typed operations over a canonical client specification: derive the package, check signal coverage, explain blockers, plan, simulate or apply, and verify.
What I built
02 / How I build AI
How I know it works
Golden datasets come from real investigations. Every prompt change runs regression tests, and every run is traced and replayable. A spent budget is reported as partial or blocked, never as success.
golden-setregressionjudgegroundinginjectiontrace03 / Experience & leadership
From applied mathematics to industrial data and AI leadership. I connect the architecture, the team, and the business problem.
Full experience & résuméThe foundation
Optimization, probability, statistics, and graph theory — a foundation for understanding models, not just using them. My thesis: Denoising Diffusion Models and Sequential Monte Carlo methods for image inpainting.
Education & qualifications →Professional certifications →Sep 2025 — Present
Narrativewave
Own AI strategy, architecture and delivery across three connected products, and lead a team of five AI engineers and data scientists.
Team of 5 · 3 AI products · weekly sessions with 3+ enterprise clients
Sep 2023 — Sep 2025
Narrativewave
Built the company's standard analytics packages and owned enterprise delivery across 12 GW+ of installed capacity, bridging customers, Product, and engineering.
12 GW+ delivery · Solar, Wind & BESS packages
Feb 2022 — Sep 2023
Narrativewave
Owned end-to-end delivery for three customer accounts, primarily Oil & Gas: asset onboarding, signal mapping, data validation, predictive models, and field feedback.
3 accounts · millions of sensor readings a day
Nov 2020 — Feb 2022
Crypto Trading Bots · Independent project
An independent project connecting applied mathematics to automated trading: ML signals, backtesting, and risk management.
Independent project · ML signals and backtesting
Mar 2021 — Jan 2022
National Institute of Genomic Medicine (INMEGEN)
Applied machine learning and statistical modeling to genomic data, building a foundation in complex scientific datasets.
Research internship · genomic data
04 / Beyond the day job
Not every project starts with a brief.
Sometimes, I just want to see it exist.
Personal project · Open source
A pixel-art office for the AI coding agents running on your machine.
Projects become floors. Conversations become coworkers. Under the tiny desks: Rust, local-first integration, and deliberate boundaries between watching work and controlling it.
Development preview · not a published release