Climb.Analyze.Build.

Business Administration graduate, University of Turin. I build AI products and climb 300-metre walls. Sport discipline and business rigor, in one place.

When I have a goal and the confidence to reach it, I don't stop. Running taught me to push through the hard parts. Climbing taught me to read risk calmly and keep fear out of the decision. I bring both to business: decide on numbers, own the result, build things people use.

How I work

Four rules I apply on a wall and in a spreadsheet.

Extreme ownership

Full responsibility for results, no excuses.

Data-driven mindset

I decide on numbers, not impressions.

Results focus

I measure the value of every action by its impact.

Calculated risk

I weigh every move before I make it.

What I bring

Economics: Controlling, accounting, financing, forecasting, strategy, marketing, social media strategy
Sales: Outbound prospecting and cold outreach on LinkedIn, BDR/SDR pipeline building, demo booking and show-rate management, lead qualification, discovery calls, objection handling, CRM operations in Attio, sales process and website improvement. At Legis: 200+ people contacted, 100+ demos booked in one month at a 60% show rate, outbound clients worth €45k in revenue
Languages: Italian (native), English (B2)
AI: Agentic AI and vibe coding: Quotify, B LAB
Communication: Public speaking, Instagram, TikTok

Projects

Not in production

Quotify

AI quote generation for Italian manufacturing SMEs. A B2B SaaS that turns quoting and analytics into one tool.

Case study

Not in production

B LAB

An AI climbing coach. Computer vision reads how you move and gives biomechanical feedback on every attempt.

Case study

Thesis, 2026

AI agents & business

Research on how autonomous AI agents are reshaping companies.

Read the thesis

Essay

Controlling essay

Why cost controlling is the reality check of a business, with a worked cost-center allocation.

Read the essay

2022

Two e-commerce stores

Launched and grown organically: brand development, photography, influencer management and content for TikTok and Instagram.

Founder

In motion

Simone climbing an overhanging rock face in a forest

Simone skiing with the Matterhorn behind him

Simone on a multi-pitch route above a valley

Simone bouldering on a large boulder in a valley

Simone climbing a steep rock face in the woods

Post-marathon stretching with run stats: 42.01 km in 4:29:08

Experience

Oct 2026

Founder Associate, then SDRLegis srl, AI tools for DPOs and compliance

BDR and SDR work on LinkedIn: 200+ people contacted, 100+ demos booked in one month at a 60% show rate. Improved sales operations with Attio and the company website.

2025, 2 months

Curricular internPalazzo Rosa, Barbania

Export management (Kuwait, Lithuania, Switzerland), marketing, AI photo enhancement, marketplace optimization on Alibaba, Faire and Ankorstore.

2023 to 2025

Dining room managerCascina Margherita, Barbania

Premium client management, dish presentation and dining room coordination at a restaurant aiming for a Michelin star.

2022

E-commerce founder

Two online stores built from zero.

Education

2023 to 2026

BSc Business AdministrationUniversity of Turin

Controlling, accounting, marketing, social media management, finance. Graduated June 2026.

2018 to 2023

Scientific high school diplomaLiceo Scientifico Aldo Moro, Rivarolo Canavese

Where I break it down

Instagram@Ruffino_simone
Startup terms explained TikTok@ruffino_simone
CAC, LTV, MVP YouTube@Simoneruffino
AI agents, explained

Building a quoting agent from scratch

How I designed and shipped Quotify, and what validation taught me.

Read the case study

An AI coach that never launched

Three months on B LAB, and the market numbers that ended it.

Read the case study

AI agents and business

My research on how autonomous AI agents are reshaping companies.

Read the thesis

The engine room of business

Why controlling is the ultimate reality check.

Read the essay

All projects

Quotify

An AI agent for automated B2B quoting.

Year: 2026
Type: B2B SaaS, business intelligence, AI agent
Built with: A business partner
Status: Not in production

Quotify overview dashboard — Overview dashboard: revenue, quotes issued, close rate and a live feed of the agent’s activity.

The problem

Many small and medium businesses keep three or four people on administrative work that is mostly quotes: read the request, check the price list, draft the document. That is a fixed cost. Every manual step is also a chance for an error that loses the sale, and slow replies lose customers.

Quotify turns that fixed cost into a variable one. An AI model working from the owner’s own pricing logic drafts quotes at any hour, and a person reviews each one before it goes out.

Where it came from

I conceived it in April 2026 as a university course project for an economics presentation. It was picked over a competitor-monitoring idea because it could be demonstrated live. A two-week MVP deadline was set in May 2026, and I built it with a business partner.

Three modules

Executive dashboard

Revenue trend, quotes issued, close rate, average deal size, and a live feed of what the agent is doing in the background.

AI quoting engine

Reads incoming emails, extracts the data and drafts a quote from the price list and rules. The owner adjusts discounts or line items and authorizes sending. A human stays in the loop; it is not autonomous.

Receivables and cash flow

Tracks payment deadlines, flags overdue invoices and runs reminder sequences with an adjustable tone.

Setup in about three minutes

Connect Gmail, upload a price list, configure the agent. Settings cover tone, behavior rules and hard limits on discounts and pricing.

How it was built

Built lean, with iteration speed ahead of early architecture. Much of the implementation was done with Claude Code, which cut the time and cost of building from scratch.

Stack: Next.js 16, Supabase, Drizzle ORM, Claude API, Inngest, Resend, React PDF, Git and GitHub, Cursor.

What validation showed

The engine works best when pricing is rigidly parametric: price times surface area for cleaning or landscaping, or daily rate times days times units for equipment rental.

Interviews with a healthcare provider and two HVAC and plumbing companies showed the limits. Plumbing quotes depend on a site visit and unpredictable labor, which a multiplier engine cannot handle. Non-technical owners would also need a far simpler setup interface.

What I learned

A working product is not adoption

The real question was whether an owner in his fifties would set up an API key and import a price list. Onboarding and usability matter more than extra features.

A business founder can ship an MVP alone with AI tools

Scaling, complex databases and data security still need a technical co-founder.

You cannot design well for a market you have not worked in

Without field experience you miss the everyday frictions and exceptions that run those businesses.

Screens

Quotify onboarding wizard — Onboarding: connect Gmail, upload a price list, configure the agent.

Quotify agent chat — Chat with the agent to draft quotes, explain rates or summarize pending clients.

Quotify agent settings — Agent settings: tone, behavior rules and hard limits on discounts and pricing.

Quotify product catalog — The price catalog the agent quotes from.

Quotify receivables view — Receivables: overdue, due soon and total portfolio, with one-click reminders.

Quotify generated quote — A branded, itemized quote, ready to send or download as PDF.

All names, amounts and contact details in these screens are demo data.

Full article Source code on GitHub

Next projectB LAB

All projects

B LAB

An AI coach that teaches you to climb.

Year: 2025 to 2026
Duration: Three months
Type: Computer vision, YOLOv8, product design
Status: Shut down after market validation

B LAB system architecture illustration — System architecture: video upload, pose and hold detection, biomechanics engine and an AI coaching layer. Illustration.

In numbers

17body keypoints tracked in real time

5 mmaximum wall height in the indoor setting

7,000+images of set walls used to train hold recognition

200,000climbers in the real market, against about 1 million guessed

The 200,000 figure comes from official FASI registration data.

The problem

In bouldering, technique counts for more than raw strength. I started indoors: walls capped at 5 metres, with crash mats, allow safe repeated attempts and lighting you can control.

The first idea

It began as a “beta solver”: photograph a wall, detect the holds and generate a move sequence shown on a stylized avatar. Behind it sat an assumption I never tested, that other climbers share my obsession with performance.

How it worked

The pipeline: video upload, pose and hold detection, a biomechanics engine, then an AI coaching layer. In March 2026 I dropped fixed thresholds for judging movement: computer vision only measures, and a language model acts as the coach.

Pose estimation

A YOLO-based model tracks posture, joint orientation and center of mass.

Hold recognition

A second model locates every hold on the wall.

Efficiency traffic light

The skeleton and the hold map combine into a biomechanical analysis. Green is efficient, yellow is uncertain, red is inefficient, with an error log after each attempt.

Where the engineering broke

The rigid efficiency rules flagged deliberate, correct techniques as errors: dynamic moves, cut-loose moves, asymmetric loading. A single camera also had no depth, so it could not measure foot pressure or how far the hips sit from the wall.

Why it stopped

The validation trap

Community feedback, including Reddit discussions, showed climbers often resist technology in a sport they love for nature and social contact. The problem was not felt as a problem.

Willingness to pay

At about 15 euros a month, the subscription met resistance. Climbers did not want another recurring fee.

Market size

I had guessed about a million practitioners. Official registrations pointed to about 200,000, too few for a scalable model, and break-even looked out of reach.

What I took from it

Technical

Faster development with AI coding tools, full-stack ownership from a Python backend to version control, and serverless infrastructure with a relational database for users and a waitlist.

Strategic

Focus on the problem, not the solution. Validate demand before building complex systems, research community channels and competitor history early, and decide on market data instead of intuition.

Stack

Python, YOLOv8, Supabase, Cursor, Claude, Netlify, Hostinger, Git and GitHub.

Brand and landing page

B LAB waitlist landing page — The waitlist landing page for B LAB.

B LAB manifesto — The manifesto section of the landing page.

Full article Frontend repository Backend repository

Next projectAI agents and business

All projects

AI agents and business

What I learned exploring the economic impact of AI agents.

Where: University of Turin, Business Economics
Academic year: 2025/2026
Advisor: Prof. Giacomo Büchi
Approach: Economic analysis, not technical

Thesis cover — Cover of the thesis. Illustration.

Numbers in the thesis

80 to 95%of AI projects estimated to miss their expected value

25%of hours in administrative, legal and engineering work estimated automatable

71%of large Italian companies have started AI projects

8%of small Italian businesses have

Figures as cited in the thesis. It does not tie each one to a single source.

Starting point

My interest began when an AI system reasoned through a cost-analysis problem in front of me. That connected the technology to what I was studying in economics.

The argument

From CAPEX to OPEX

Building your own AI infrastructure gives control and data security, at the price of high fixed costs and the risk of obsolescence. Cloud and API models turn that spending into OPEX and make it elastic, but without strict governance, consumption-based costs can get out of hand quickly.

The true ROI

Real advantage comes from redesigning the business model, not from small efficiency gains. The hidden costs are data breaches, regulatory penalties, reputational damage and hallucinations, which is why human validation stays necessary.

The labor market paradox

The likely result is polarization, not wholesale replacement. Specialized skills earn a wage premium while routine intermediate tasks lose bargaining power. Complementarity and meaningful human oversight, as the EU AI Act frames it, are the way forward.

The Italian gap

Large companies are far ahead of small ones. For SMEs the main barriers beyond cost are a shortage of qualified people and immature data management.

Conclusion

AI is a strong deflationary and transformative force, but not a magic solution. Continuous reskilling and upskilling are the only way to stay relevant.

Sources and tools

Sources named in the piece: McKinsey, Stanford HAI, Bank of Italy and the Politecnico di Milano AI Observatory. I used Claude as a research and writing assistant.

Full article

Next projectThe engine room of business

Let's connect.

LinkedInBest way to reach me Instagram@Ruffino_simone TikTok@ruffino_simone YouTube@Simoneruffino

Climb.Analyze.Build.

How I work

Extreme ownership

Data-driven mindset

Results focus

Calculated risk

What I bring

Projects

Quotify

B LAB

AI agents & business

Controlling essay

Two e-commerce stores

In motion

Experience

Founder Associate, then SDRLegis srl, AI tools for DPOs and compliance

Curricular internPalazzo Rosa, Barbania

Dining room managerCascina Margherita, Barbania

E-commerce founder

Education

BSc Business AdministrationUniversity of Turin

Scientific high school diplomaLiceo Scientifico Aldo Moro, Rivarolo Canavese

Where I break it down

Building a quoting agent from scratch

An AI coach that never launched

AI agents and business

The engine room of business

Quotify

The problem

Where it came from

Three modules

Executive dashboard

AI quoting engine

Receivables and cash flow

Setup in about three minutes

How it was built

What validation showed

What I learned

A working product is not adoption

A business founder can ship an MVP alone with AI tools

You cannot design well for a market you have not worked in

Screens

Read more

B LAB

In numbers

The problem

The first idea

How it worked

Pose estimation

Hold recognition

Efficiency traffic light

Where the engineering broke

Why it stopped

The validation trap

Willingness to pay

Market size

What I took from it

Technical

Strategic

Stack

Brand and landing page

Read more

AI agents and business

Numbers in the thesis

Starting point

The argument

From CAPEX to OPEX

The true ROI

The labor market paradox

The Italian gap

Conclusion

Sources and tools

Read more

The engine room of business

The argument

Operations come first

Cost analysis shows the real economics

Entrepreneurs need granular cost visibility

Unit economics as the default lens

Rigor turns intuition into evidence