Introducing Vasuki AI: Building Our Custom Conversational Model from the Ground Up
Artificial Intelligence is shaping the future of how humans interact with machines. While powerful large language models exist, building your own AI assistant — one tailored to your company, your users, and your unique data — is a whole different challenge.
Meet Vasuki AI, our custom conversational model, now live at vasukiai.netlify.app. It’s a step toward the next generation of company-specific, developer-friendly AI models.
The Vision Behind Vasuki AI
When we started, our goals were clear:
- Domain Awareness: We wanted an AI that knows about our organization, roles, and products — not just general knowledge.
- Fast Response: No bloated infrastructure. It had to be lightweight and respond in real-time.
- Customizability: Something we could tweak and improve without depending on a black-box external model.
- Developer-Friendly Integration: Web-deployable, modular, and easy to embed into other apps.
We envisioned Vasuki AI not as a competitor to massive LLMs, but as a purpose-built conversational engine — small, focused, and constantly evolving.
What Vasuki AI Can Do
Even in its current version, Vasuki AI is showing how custom conversational models can bring value:
Mathematical Problem Solving
solve 90938**21
➡️ The result is 1.360358483096574e+104.
Scientific Explanation
why do plants require sunlight
➡️ Plants require sunlight because light energy is absorbed by chlorophyll molecules in their leaves and converted into chemical energy during the process of photosynthesis, which produces glucose for growth and oxygen as a byproduct essential for life on earth.
Organizational Knowledge
who is CEO of Vasuki
➡️ my developer and owner, more importantly - the CEO of Vasuki iTech is Adamya Khairwal. an inspiring 14 year old leader who built me with his proficiency in javascript.
Technology Basics
what is artificial intelligence
➡️ Artificial intelligence is the simulation of human intelligence in machines that are programmed to think and learn.
Machine Learning Fundamentals
what is supervised learning
➡️ A supervised algorithm is a machine learning algorithm that requires labeled input data for training.
This is just the starting point. The more we integrate domain-specific data, the smarter it gets.
Technical Deep Dive
Here’s what’s happening under the hood:
- Knowledge Base: Curated structured and unstructured data covering company information, technical definitions, and domain facts.
- Query Processing: Light tokenization and keyword extraction to understand the user input.
- Scoring System: Each candidate response is scored based on similarity (TF-IDF and cosine similarity in this version) with a confidence value attached.
- Hosted Frontend: A simple, modern web UI hosted via Netlify, making it accessible instantly.
- Scalable Design: While the MVP uses JavaScript-based scoring, we designed the pipeline so we can swap out or layer in neural inference models in future versions.
Why We Built Instead of Bought
Building from scratch wasn’t just a passion project — it was strategic.
- Cost Control: No per-token or API call fees.
- Data Privacy: Full control over what data is stored, processed, or logged.
- Customization: From role names (like our CIO, Jatin) to business-specific definitions — no third-party fine-tuning needed.
- Future Proofing: A foundation we can improve, integrate with internal APIs, and eventually hybridize with external models when needed.
Try It Yourself
We’re sharing the live preview of the current Vasuki AI model here:
👉 vasukiai.netlify.app
You’ll see it respond to a variety of topics — some general, some company-specific — with a confidence score to indicate certainty.
What’s Next
This is just Version 1. Upcoming improvements include:
- Context Memory: Keeping track of previous questions for multi-turn conversations.
- Hybrid Search: Combining symbolic reasoning with semantic search for more accurate responses.
- Neural Upgrade: Adding lightweight neural inference models for natural language generation.
- Analytics: Monitoring question patterns to identify gaps and retrain efficiently.
Final Thoughts
Vasuki AI represents what’s possible when you combine focus, modular design, and a vision for tailored AI systems. Instead of waiting for big tech to solve every problem, we can build our own — smaller, faster, domain-specific assistants that give companies an edge.
We’re just getting started. 🚀
Check it out, break it, test it — we’d love your feedback:
👉 vasukiai.netlify.app
Top comments (0)