Visual Reasoning AI
Until now.
Drik Vision AI
A vehicle intelligence layer over any camera feed. Officers ask in plain language — Drik returns evidence frames, map, timeline, and graph. No GPU upgrades. No retraining. Just plug in and ask.
Seven Atomic Capabilities
Officers ask natural queries. Drik returns evidence frames, map, timeline, and graph. The same seven primitives serve army, border, police, and traffic.
Every historical sighting across every feed.
From last sighting + direction + speed.
Dominant overnight zone.
Repeated-path corridor.
Vehicles that repeatedly co-appear.
"White Bolero, dented bumper, roof carrier" — ranked results.
Helmet-less, red-light, wrong-way, overspeed (cross-camera), triple-ride, seatbelt, phone, tint, plate tampering, no-parking, school-zone.
Capability 7 is the self-funding wedge. Capabilities 1–6 are the intelligence depth.
Four Verticals
The seven atomic capabilities compose into 18 deployment scenarios across army, border, police, and traffic.
Scenarios 1–6
Convoy, cantonment, and counter-intel work that a radio race cannot keep up with.
Scenarios 7–9
Infiltration corridors, handler–asset relationships, and behavioural signatures of smuggling.
Scenarios 10–15
72 hours of 12 feeds compressed to minutes. Network discovery without prior intel.
Scenarios 16–18
Zero capex to the state. Majority share of challan revenue stays with the buyer.
Selected Scenarios
The self-funding wedge first, then three intelligence-depth scenarios. Each composes one or more of the seven atomic capabilities above.
Eleven violation classes. Zero capex to the state. Majority share of revenue stays with the buyer — the system pays for itself from month one and funds the intelligence depth that follows.
Flag any vehicle or person of interest. The system watches every camera feed, 24/7.
The moment a watched target appears on any camera, you get an instant alert with location and confidence.
See where a vehicle is heading before it arrives. Predicted routes from historical patterns.
In-house Dataset
Bullock carts to Boleros, fire vans to e-rickshaws, sugarcane jugaads to school buses — trained on the messiest road network on earth, not on COCO.
Built by mixing traditional computer vision with modern large models. Pure LLM pipelines are too slow at camera-feed scale; pure classical detectors miss the long tail. Drik runs classical perception at the edge and uses large models surgically — for description, refinement, and ambiguity resolution.
How It Works
Deploy Drik on-premise, in the cloud, or both. Our nodes scale with your camera network — from a single building to an entire city.
Five Levels of Visual Reasoning
It sees every object
Every vehicle, every person, every object — detected, classified, and counted in real time across every frame.
It reads every plate, every face
License plates extracted. Vehicle makes and models identified. Colors, modifications, damage — all catalogued instantly.
It describes what it sees in words
Not just labels. Full natural-language descriptions of every scene, every event, every anomaly. Searchable. Queryable.
It understands cause and effect
Why did traffic stop? What caused the accident? Which vehicle triggered the chain reaction? The AI builds causal chains.
It anticipates what happens next
Pattern recognition across time and space. Congestion forecasts. Accident risk zones. Before it happens, the system sees it forming.
What’s Next
Today: vehicle intelligence over any camera feed. Next: a vision foundation model that reasons about scenes the way a human brain does — not frame-by-frame detection plus classification, but a unified system that builds and updates a model of the world from camera feeds.
DVWM is research, not product. No demo, no benchmarks. The honest framing: we are building toward it because the next decade of camera AI is reasoning, and per-scenario engineering does not scale to it.
The Proving Ground
1.4 billion people. 300 million vehicles. Zero lane discipline.
Jugaads. Ox carts. Overloaded trucks. Band baarats. Cows.
Modified vehicles. No lanes. No structure.
If it reasons here, it reasons everywhere.
The People Behind Drik
Forbes India 30 Under 30 (2025). IIT Guwahati alumnus building what cameras were always meant to be — systems that truly see, reason, and act.
"India has 50 million CCTV cameras and almost none of them think. I want to change that."
Track Record
Backed By
Forbes India 30 Under 30
2025 · Recognized among India's top young innovators
TIDF · IIT Guwahati
Incubated & backed by DST, Govt of India
Govt of Assam
Backed by the State Government
Industry Traction
LoIs from government offices · ITC & system integrator partnerships
The Team
Hansraj Patel
Founder & CEO
Computer Vision Engineer
Full-time · Remote · India
Full-Stack Engineer
Full-time · Remote · India