In this Rise of AI Virtual Chat recorded live in April 2020, Fabian Westerheide interviews Aleksandar Stojanovic, chairman and co-founder of AVA, an AI company he started in 2014 after earlier careers in music production and management consulting. Stojanovic describes AVA as a 'fusion platform' that pulls in social media, local and international news, professional intelligence feeds and sensor data, standardizes it, and uses machine learning plus human labeling to give organizations a contextualized, real-time picture of security risk, rather than predicting individual crimes. He walks through clients including DHL, the London Metropolitan Police and Deutsche Bahn, a live map demo of historical incidents in New York, and how AVA combines organizational, epidemiological and socio-political data to help multinational companies assess pandemic risk, citing the Canadian company BlueDot's early COVID-19 detection through the ProMED physician network. He also addresses privacy, arguing AVA, built after GDPR, deliberately avoids personal profiling and instead aims to democratize access to security intelligence historically available only to the wealthy.
Key takeaways
- AVA aggregates social media, news, professional intelligence and sensor data into one common standard, using both machine learning and 24/7 human labeling teams.
- Stojanovic says AVA's clients include DHL's global security operations, the London Metropolitan Police and City Hall, and Deutsche Bahn.
- He cites BlueDot, a Canadian company using the ProMED physician network, as having detected COVID-19 roughly one to two days before the World Health Organization.
- Assessing pandemic risk for reopening an office requires combining organizational, epidemiological and socio-political data, none of which alone is sufficient.
- Stojanovic says AVA deliberately avoids building personal profiles of individuals, focusing on aggregate risk context rather than targeted surveillance.
Detailed notes
From music producer to security-AI founder
Stojanovic recounts starting in the music and entertainment business in the 1990s, moving into project management consulting where he built his own consultancy of 40 to 50 people, then selling that business and founding AVA in 2014, the company he now chairs and co-founded.
What AVA is: a fusion platform for security intelligence
Stojanovic says most people see AVA only through its interfaces, operator dashboards, mobile apps, or API integrations, but the platform underneath is a 'fusion' system that ingests reports from social media, local through international news, paid professional intelligence providers, and sensors such as smoke alarms, and translates all of it into one common data standard. A training team and a small human 'monitoring' team work around the clock, especially to label ambiguous cases and avoid automated action on severe incidents, feeding a machine-learning loop that Stojanovic says is now good at automatically detecting things like fires, explosions and shootings, though some categories remain under-trained.
Clients: DHL, London police and Deutsche Bahn
Stojanovic describes AVA's relationship with DHL's global security operations team as a close partnership, and says London's Metropolitan Police and City Hall use the platform to contextualize organized and street crime for crime prevention. Deutsche Bahn uses AVA to maintain situational awareness across the country, with a stated goal of eventually pushing alerts, for example about hooligans traveling to a soccer game, directly to individual train station staff. Stojanovic says organizations without dedicated security teams, such as police forces or public transport operators, typically start by lacking any systematic data at all, while organizations like DHL already have large internal intelligence capability but struggle to filter thousands of daily data points to find what is relevant to each analyst.
Predicting pandemic risk across three layers
Discussing COVID-19, Stojanovic cites the Canadian company BlueDot as having detected the outbreak roughly one to two days before the World Health Organization, using the ProMED network of physicians worldwide. He says assessing an organization's pandemic risk requires combining three layers: internal organizational data (infection cases among staff, supply chain and site-attendance data), epidemiological data on infection spread and healthcare capacity, and a socio-political layer covering civil compliance, government responses such as curfews or mask rules, and unrest risk, which he notes the UN Secretary-General had recently flagged as a global concern. He says no single existing platform currently combines all three layers, which is the gap AVA aims to fill for decisions such as when to reopen an office.
Privacy, GDPR and democratizing security intelligence
Stojanovic says AVA was built after the EU's GDPR was already being drafted, so compliance was designed in from the start, and that the platform deliberately does not build personal profiles or track named individuals, distinguishing its aggregate risk-context approach from targeted policing systems. He frames the value as democratizing security intelligence that was previously available only to wealthy individuals and organizations with private security teams, comparing the experience to a medical blood test: uncomfortable information that nonetheless allows people to act and improve their situation.
Q&A highlights
Demonstrating the platform on a New York map, Stojanovic shows historical incident data with trust scores per source, adjustable by a human user, and filters such as searching only for robberies within a neighborhood; he describes wanting similar security-context filters eventually built into services like booking platforms. Asked about a striking pattern, he cites research finding that a 10-point rise in a city's air quality index correlates with about a 1% rise in petty crime, noting causation is still being studied. On end-user access, he says AVA is primarily B2B and reaches individuals mainly by embedding into partner apps and mobile network operators rather than through its own consumer app. Looking ten years out, he predicts the security market will become as heavily regulated and technology-driven as fintech and healthtech became before it, and says he expects AVA to become a widely used, 'world-defining' company.
Quotes
The idea was to create something like this really for everybody, so that you as an end user can see the things that are truly relevant for you.
Aleksandar Stojanovic
Nobody ever said we need the personal data and we need personal profiles.
Aleksandar Stojanovic
Questions answered in this talk
What is AVA and how does it assess security risk?
According to Aleksandar Stojanovic at a Rise of AI Virtual Chat in 2020, AVA is a fusion platform that combines social media, news, professional intelligence feeds and sensor data into one standardized format, using machine learning and human labeling to give organizations real-time, contextualized situational awareness rather than predicting individual crimes.
How did AI help detect COVID-19 early?
According to Aleksandar Stojanovic at a Rise of AI Virtual Chat in 2020, the Canadian company BlueDot detected the COVID-19 outbreak roughly one to two days before the World Health Organization by monitoring the ProMED network of physicians worldwide.
Does AVA's security AI build profiles of individual people?
According to Aleksandar Stojanovic at a Rise of AI Virtual Chat in 2020, AVA deliberately does not build personal profiles of named individuals, focusing instead on aggregate, contextualized risk data, a design choice made in compliance with GDPR from the company's founding.