Simone Abbiati Portfolio

1 · Product understanding

I understand the product before I market it

A PhD in computational linguistics, and the habit of reading the repository before writing the message. I work directly with engineering and product: I build with the same tools, ask about the metric and the baseline, and only then write.

github.com/SimoneabbiatiProfile ↗

461 contributions in the last year63 active days199 commitsless more

LanguageAppShapeling's back end: FastAPI, async PostgreSQL with pgvector, Claude, speech-to-text and voice, a CEFR difficulty gate in spaCy.Python · private

LanguageLearnerProfilingA voice-first AI interview that profiles a language learner in under seven minutes.Python · private

squirro-demo-voiceoverffmpeg, Claude vision and ElevenLabs: silent product demos become narrated videos.Python · private

KG-CourseAssignments on Squirro's Graphite knowledge-graph API.Jupyter · private

LaScalaNext.js and Prisma app for the stage managers of Teatro alla Scala.TypeScript · public ↗Lit_reviewScripts for collecting and processing literature for systematic reviews.Python · public ↗

Academic research

PhD in computational linguistics

University of Bergamo, with visiting periods at Stanford (LLMs for information retrieval) and Cambridge (AI ethics). Research on AI-generated text, disinformation and how language shapes trust.

All publications on Google Scholar ↗

461GitHub contributions, last 12 months

16repositories

4 + 3peer-reviewed papers, three more under review

2 · Positioning & the buyer

Then I map who I am talking to

Once I understand the product, I map the buyer: a 223-term persona taxonomy in Graphite, then positioning, ICP and personas built on it. Drag the map; click a node to follow a buyer to their pains and the use cases that solve them.

What the graph is for. It refines product marketing positioning: every message is checked against a real buyer, pain and use case. And it automates content production with AI: each buyer × pain × use-case combination becomes a brief that an LLM turns into a first draft, reviewed before it goes out.

Simplified for the portfolio: links show typical relations, not client data.

223taxonomy terms

46buyer roles

30validation interviews planned

3 · Go-to-market & content

Then I take it to market, in every format

Every launch goes out in many formats: white papers, carousels, videos, essays, webinars, newsletters and nurture emails. I re-cut each piece for the persona I am talking to, the same product told in the terms that make it matter to that buyer, and the webinars and emails turned attention into leads. The actual covers are below.

Why knowledge graphs make enterprise AI more accurateLaunch campaign

AI for Good: two universities, one research programme

A research programme on GenAI in educationPartnership launch

How knowledge graphs and GenAI work together in searchWhite paper

Knowledge graphs, explained in one city walk

Knowledge graphs explained through a walk in ZurichExplainer video

Stop hunting for the right cell. Finally model.

Less data hunting, more modelling for financial analystsProduct carousel

One AI research paper, made useful for businessCarousel series

The key findings of Stanford's AI IndexReport summary

Why chunk-based RAG retrieves the wrong contextCarousel series

A research paper on GenAI and knowledge managementPaper explainer videos

How a smaller AI company competes, from DavosShort video

Twelve years of company history, told by its peopleEmployer brand

AI as an educator: sleepwalking into a nightmare or dancing on the horizon of a dream?

The promise and risks of AI in educationThought-leadership essay

GenAI and the European Health Data SpaceWebinar

GenAI in education: an open conversation

A panel with two universities on GenAI in teachingWebinar

How AI speeds up employee onboardingCampaign

LinkedIn analytics · last 12 months

+59%impressions, year on year

21K+impressions on my posts

48%of views from outside my network

33%of my audience at senior level

12content formats

+187%Academy subscribers

+23%marketing-qualified leads

Launch campaignKnowledge graphs for the enterpriseAfter an acquisition brought knowledge-graph technology in, a campaign to explain it: white paper, explainer video, short videos, posts and an email to the Academy audience.My role: Content lead · Campaign on request
Go-to-market planGenerative AI for regulated industriesA plan for finance and the public sector: a persona schema for compliance buyers, a compliance proof pack and an end-user certification for audit-ready use.My role: Author · Summary on request
Partnership launchAI for Good: two universities, one research programmeA collaboration between Liverpool Hope University and Universidad del Norte on generative AI in education, launched with a joint webinar.My role: Brokered it, ran the launch · Campaign on request
White paperKnowledge Graphs for EnterprisesHow knowledge graphs and generative AI combine in enterprise search, with use cases by industry.My role: Wrote it, briefed the visuals · PDF on request
Explainer videoKnowledge graphs, explained in one city walkTwo minutes in Zurich: landmarks as nodes, streets as relations.My role: Script, voice, on camera · Watch on YouTube ↗
Product carouselStop hunting for the right cell. Finally model.A carousel for financial analysts that sells the outcome, more time modelling, before the product.My role: Copy, concept · Samples on request
Carousel seriesSmart Reads, Smarter MovesOne research paper per carousel, rewritten for people who make business decisions.My role: Concept, copy, layout · Samples on request
Report summaryThe 2025 AI Index Report, in slidesThe Stanford AI Index cut down to what business leaders need from it.My role: Copy, layout · Samples on request
Carousel seriesSlide-sized NLP: the trouble with RAGA technical series for engineers: why chunk-based retrieval overloads context, and what better retrieval looks like.My role: Concept, copy · Samples on request
Paper explainer videosOne paper a month, explained on cameraShort videos that walk through a research paper on knowledge management and generative AI.My role: Script, on camera · Samples on request
Short videoIn a world of giantsA short social video from Davos during the World Economic Forum week, on how a smaller AI company competes.My role: Script, on camera · Video on request
Employer brandThe Squirro story, 2012 to 2024Company history told through the people who lived it: a course, a timeline and a carousel.My role: Interviews, copy · Samples on request
Thought-leadership essayAI as an educator: sleepwalking into a nightmare or dancing on the horizon of a dream?An essay that sells the point of view before the product.My role: Author · Read on squirro.com ↗
WebinarLLMs and knowledge graphs in healthcareA webinar on the European Health Data Space: agenda, guest, promotion, hosting. Fixed format of 20 minutes talk, 15 interview, 10 questions, with reminder emails.My role: Producer and host · Recording on request
WebinarGenAI in education: an open conversationA panel with Liverpool Hope University and Universidad del Norte, promoted with both universities' marketing teams.My role: Organiser, moderator · Recording on request
CampaignAI for employee onboardingA carousel that turns an Academy course into a question every HR and operations lead asks.My role: Copy, course tie-in · Samples on request
Email programmeAcademy newsletter and nurture emailsA newsletter template, a nurture sequence for prospects, webinar invitations and personalised HubSpot announcements.My role: Copy, segmentation · Samples on request
LinkedIn programmeA weekly editorial rhythmTwo to three posts and a video a week, then a monthly plan: one paper video, one carousel a week, three posts a week on Tuesday to Thursday.My role: Planning, writing · Plan on request

4 · On stage

And I say it out loud, to four kinds of audience

Technical and corporate audiences at AI conferences, institutions and policy makers with the Bassetti Foundation, buyers in finance and industry, and universities.

See the post ↗

Four kinds of audience

Technical & corporateAI conferences and industry forums: GenAI London, GenAI Zurich, PanoramAI, Virtuelle Fabrik

Institutions & policyRegione Lombardia, the Swiss Consulate General in Milan and the Bassetti Foundation

CommercialFinance buyers in Athens and an industrial client in Bergamo

AcademicConferences around the world during my PhD, St. Gallen and university partnerships

100+people per audience at my talks, in English and Italian

Technical & corporate

AI conferences and industry forums

Oct 2025GenAI LondonTeaching GenAI to understand: classification and knowledge-graph enrichment ↗

Jun 2025PanoramAI Summit, ZurichNavigating startup challenges in the post-ChatGPT AI landscape ↗

Oct 2024Virtuelle FabrikLeveraging AI for Industry 4.0

May 2024GenAI ZurichTransforming healthcare with GenAI and the European Health Data Space

Academic

Universities and research

Sep 2024Mobile Business Konferenz, University of St. GallenCompeting in the AI revolution

Oct 2024Liverpool Hope & Universidad del NorteGenAI in education: a panel webinar I organised and moderated

2019–2023Conferences worldwidePapers at linguistics and AI conferences during my PhD

Institutions & policy

With the Bassetti Foundation and public institutions

Feb 2026PoliS Lombardia, Regione LombardiaAI and sustainability, with the Bassetti Foundation ↗

2026MEET, MilanItaly–Switzerland dialogue on responsible innovation, with the Swiss Consulate General

2023–nowBassetti FoundationAdviser on AI and public communication, with legislators and academics

Commercial

Buyers in finance and industry

Nov 2024GenAI Summit, AthensTransforming financial insights: AI for risk management and ESG analysis ↗

2026Global industrial manufacturer, BergamoA hands-on use-case workshop that led to a pilot

100+people per audience at my talks

4kinds of audience

13%MQL-to-SQL conversion from people who contacted me after my talks

5 · Sales & partner enablement

Material sales and partners actually use

Marketing comes first; then I make sure the people who sell can use it. I equip sales with courses, pitches and demos that produce themselves, and partners, who deliver much of the work, with a learning plan and a dashboard that shows how far each one has got.

Partner adoption dashboard with company names blurred — **Partner adoption dashboard** · certification progress for every partner organisation, built on Graphite. Partner names blurred.

Product demo of the RFP agent — Demo film

Also in this card

Automated demo generator
Partner lifecycle learning plan
Partner enablement spaces
Sales, pre-sales and pricing courses

594partner and delivery staff tracked

10learning paths: 61 courses, 25 lessons

Autonarrated product demos

6 · Customer enablement

Customers who know how to use the product

The same enablement mindset, applied to customers: 10 learning paths with 62 courses, 35+ of them migrated on learn.squirro.com, and certifications that show who can really use the product.

**Learner dashboard**XP, levels, badges and a leaderboard on learn.squirro.com

**Courses**Interactive lessons, such as LLM-powered agents in banking

**Certification**Implementation Guide modules, graded automatically

Also in this card

User guides For a GenAI search and chat rollout
Live sessions Webinars and office hours
Feedback loops 1-1 interviews, fixes shipped
Upsell Education as a way in

Learning platformlearn.squirro.comThe public training and certification platform, built solo to replace a third-party LMS: information architecture, copy, front end, certification flow.My role: Built it end to end · Visit the platform ↗
Learning paths10 learning paths by role and by solutionEach path bundles many courses, for engineers, partners and business users, plus internal paths on pricing, sales and the company story.My role: Curriculum, writing · Outline on request
Certification programmeAn 8-module partner certification with automatic gradingPartners work in their own instance; a grader checks the real configuration and issues the certificate.My role: Designed programme and grader · Walkthrough on request
Customer enablementUser guides for a GenAI search and chat rolloutFor a medical-research customer: videos, guides and FAQs on when to search and when to chat, how filters and languages change what is retrieved, and what the feedback buttons really do.My role: Requirements, content · Outline on request
AI topic coursesCourses on the ideas behind the productKnowledge graphs in the enterprise, autonomous agents, small language models and information retrieval, so customers understand why the product works the way it does.My role: Author · Course list on request
Live sessionsIntegration webinar and technical office hoursLive sessions where customers bring real questions, recorded and added to the course catalogue.My role: Host · Recording on request
Learner engagementGamification for the learning platformXP, levels, badges, streaks and leaderboards, designed from the events the platform already logs, to keep learners coming back.My role: Design · Design on request
Feedback loopsInterviews that changed the productOne-to-one feedback interviews with seven engineers on the Squirro CLI, plus partner and customer reviews of the new platform, turned into fixes and new courses.My role: Interviews, synthesis · Summary on request
Education as a way inCourses that open accounts and drive upsellUsing education to get into new accounts and to show existing customers what else the product can do.My role: Concept, pilot · Summary on request

7 · AI-native execution

I build with AI, not just write about it

In a small marketing team at a company growing from start-up to scale-up, AI is how I ship. I code with Claude Code, build on the Claude API, and use voice and image models to turn one idea into many formats. Every tool below I have used in real work.

Build

Claude CodeMy coding partner for apps, pipelines and dashboards

Claude APILanguage, vision and agent features inside my products

Python, FastAPI, SvelteKit, Next.jsBack ends, front ends and the learning platform

PostgreSQL + pgvector, DockerRetrieval, embeddings and deployable services

Voice & video

ElevenLabsNarration for product demos and voice for Shapeling

AssemblyAIStreaming speech-to-text for voice interviews

ffmpegCutting and assembling demo videos automatically

Research & writing

Claude, Gemini, NotebookLMReading papers, drafting, checking claims

Design

Claude Design, Canva, FigmaCourse prototypes, carousels and campaign visuals

MidjourneyIllustrations for the knowledge-graph white paper

What I built with them

Shapeling ↗Claude + AssemblyAI + ElevenLabsA voice-first language tutor, built solo end to end.

Narrated demosffmpeg + Claude vision + ElevenLabsSilent screen recordings become voiced product videos.

learn.squirro.com and the grader ↗Built with Claude CodeThe learning platform and the certification checks.

Carousels from the graphGraphite + LLMThe persona taxonomy generates LinkedIn carousel briefs.

Career CopilotClaude + voiceA local app for practising job interviews by voice.

Partner adoption dashboard ↗Graphite + Claude CodeCertification progress for every partner organisation.

6products and pipelines built

Claude Codedaily coding partner

End to endfrom repo to release

8 · Measurement, SEO & GEO

Then I measure what works, and fix it

Every format and channel gets measured: HubSpot dashboards for the funnel, search and AI-answer visibility for content, and impact dashboards for research. Then I fix what the data says is broken.

Data & reporting in HubSpot

Dashboards that show what to fix

Content only counts if the data says so. I aligned HubSpot properties and lifecycle stages with sales, taught the team to use them, and report which content, webinars and talks turn into MQLs and SQLs, and where the data is dirty.

Source of every MQL: content, webinar, talk or event
Conversion at each stage, by channel and by persona
Data-quality checks: missing lifecycle stage, duplicate companies, untagged sources

Subscriber

Lead

MQL

SQL

Lifecycle stages as defined in HubSpot. Bar widths are illustrative.

SEO

Be found

Keywords chosen with Google Trends: what people actually search, and when
Titles that carry the keyword and promise a clear answer
Headings and structure that follow the reader's question
Internal links between pillar pages and the pieces around them

GEO

Be cited by AI assistants

RAG retrieves by semantic clusters, so a whole paragraph has to lead to one concept, not just carry the keyword. That is where GEO differs from SEO
Answer-first, self-contained passages that can be lifted as they are
The same company and product names everywhere
Pages and numbers that are crawlable, not hidden in JavaScript

Why I am good at it

I know how retrieval works

I studied how language models find and rank text, and I build RAG systems myself. I know what makes a passage easy to retrieve and cite.

I make complex AI easy to understand, trust and buy.

I take enterprise AI to market, and make sure people can use it.

From the code to the customer