Announcing Pachamama 1.0
In just its first year, BotSupply has helped various clients across the globe to build Artificial Intelligent solutions which include customer support automation, chatbots etc. We’re excited to see clients using our solutions with great recall.
This encouraged us to build a complete ecosystem around Natural Language Understanding (NLU), so that our solutions can talk to clients in their native natural languages such as Danish, Italian, English, etc.
Today, as part of the first annual year, we’re announcing Pachamama 1.0 — The origins.
Pachamama 1.0 — The origins
Why this name ?
Pachamama is the goddess revered by the indigenous people of the Andes. She is the mother of earth and time. We embark this journey of ours to bring artificial intelligence to life by naming our first stable API after her. As we will grow together in this journey we will keep on changing and improving things but what ever is to follow will follow from the mother, Pachamama.
Story so far:
Not long ago, almost all of it was just a plan. And we feel proud when we look back to our first commit which had few files and now when we look at our system, well it is a “beast” in the making with an all new dashboard and bot making platform running in parallel.
We have come a long way, but we have just begun.
Why Origins ?
We plan to come up with various flavours (read versions) of each release and from our benchmark we have a lot to add on to this release of Pachamama.
Major upgrades and features:
It’s faster: Pachamama 1.0 is lighting fast! The robust framework architecture and the custom build algorithms are the key aspects which helped us to achieve 7.3x speedup on model building process and serving.
It’s flexible: Pachamama 1.0 includes various options such as Boosted NLP, Google, Wikipedia search integration, Bots, etc (as shown in Figure below). Clients can choose these features depending on the use cases.
It’s more production-ready than ever: Pachamama 1.0 promises, making it easier to pick up new features without worrying about breaking your existing models. It literally takes 5–7 minutes from uploading the client data and to having a servable API’s.
Other highlights from Pachamama 1.0:
- Fixed UTF-8 character handling when training through HTTP interface.
- Support for Unicode characters.
- Supports Danish, Italian and English languages.
- Higher accurate models.
- Highly secured API protocols.
- State-of-the-art (SotA) text classification with 94.2% accuracy.
We’re thrilled to see the adoption of Pachamama 1.0 by our clients and future clients. To hear more about Pachamama 1.0 and how it’s being used, contact us on firstname.lastname@example.org
The #WeCoCreate ecosystem continues to contribute and grow the NLU framework with new techniques like
- Episodic Memory for long term memory retention
- Tools like Embedding Projector to visualize the custom vocabularies.
- Real time feedback mechanism
- Moving closer to MPP (Massive Parallel Processing) architecture
This article was co-authored with Vishal Ranjan, who is the part of BotSupply team.