An Indian Artificial Intelligence (AI) startup is using algorithms to help the police and security agencies to identify criminals and solve cases much like a Hollywood crime thriller.
Founded in 2015, Staqu Technologies had initially set out to utilize AI in solving real-world problems. The company’s first product was a recommendations engine that now powers several online retailers in the country. But in February 2017, Atul Rai, the company's cofounder, embarked on a new mission to develop AI-powered predictive policing by using his company's proprietary image recognition, language-independent speaker identification and text processing capabilities. However, unlike what we see in the movies, it was no easy task.
Challenges in Predictive Policing
On the ground, India's police force and security agencies are engaged in the same challenges as most others, such as identifying the suspects of criminal activities while reducing the crime rate based on heuristic data.
To begin with, though, only a fraction of police records in India are stored in digital form, meaning that investigations and tracking criminals are offline and often tedious tasks. Although many of those criminal records may contain images, speech and textual data, the information cannot be extracted in real time across the region. And simple software algorithms would struggle to parse such complex data structures from the few records that are stored electronically.
Criminal Registration and Search
Staqu began tackling this issue by building from the ground up. It created a mobile app that enables police officers to record information about criminals while out on patrol. The multi-lingual app can store and compute biometric information (face, speech, fingerprints) and allows unlimited searches and indexing. The AI-powered app effectively allows officers to better identify possible suspects by allowing them to have ready access to its database of criminal records.
When I met Atul at Staqu's office in Gurgaon, a suburb of New Delhi, he animatedly explained to me their proprietary, advanced hybrid AI technology, which amalgamates different neural network models to process image, speech and text and then extract information that can be used to make better decisions.