Building for the future

BUILDING FOR THE FUTURE

We started working on AI chat products way before any of this became cool. So, we were forced to push the boundaries on building cutting edge software, algorithms, systems and infrastructure. Every single line of code at Haptik is written in house and we use zero external algorithms. We open sourced the world’s first chatbot specific Named Entity Recognition (NER) module in 2017.

OMNI CHANNEL FULL SLACK PLATFORM

CHAT-LEVEL BOT
ANALYTICS
DRAG & DROP
BOT- BUILDER
AGENT CHAT
INTERFACE

Haptik has everything you need to build and maintain great conversational interfaces. And the SDKs/plugins to work across any channel of your choice. The agent chat interface enables flexibility between going 100% bot or a combination of human-bot, and the analytics dashboard provides deep conversational sentiment analysis on what people are talking about. KNOW MORE

Comprehensive Machine Learning

Every incoming chat goes through a series of steps where the message is analyzed to determine the most suitable response. The average time to process a message is less than 1 second. Natural language processing combined with proprietary algorithms and great product design.

KEY COMPONENTS

Domain ClassifierThe first step to resolving a user’s chat is to figure out the domain the user is chatting about.

Named Entity Recognition (NER)The NER helps to detect relevant entities from user's message. Our proprietary Named Entity Recognition (NER) engine is designed ground up for chat bots and is 1st in the world to be open sourced for anyone to use. You can find the open source repository here.

Graph based chat flowsOnce we figure out the domain and entities, we figure out the intent of the conversation and the chat goes through a conversation tree to figure out the right response.

Deep Learning LayerUsing millions of data points built up through previous conversations and third party data, the machine learns to respond back to certain queries it may have never seen before and send an accurate predictable response.

ML Orchestration LayerThe ML orchestration layer takes input from all the different algorithms and picks the right response to continue the conversation. This layer helps bring everything together and makes the system more scalable. We can add and remove any algorithms as long as they follow the same protocol. This layer also helps send data to our date lake which powers our analytics engine.

Book me a Flight

LEARN MORE

Latest from our tech blog

SEE ALL