Unlock insights with HeavyIQ LLM Analytics Platform
HEAVY.AI, innovator in advanced analytics, today announced HeavyIQ a Large Language Model (LLM) capabilities to the GPU-accelerated HEAVY.AI analytics platform, enabling organizations to interact with their data through conversational analytics.
Users can easily explore their data with natural language questions and quickly generate advanced visualizations of that data. This streamlined process reduces the friction of traditional business analytics, allowing more users to quickly and effectively uncover insights.
“In the past, most analytics solutions required users to write their own SQL queries or learn complex drag and drop interfaces that required dozens of clicks. They would also have to bring in engineering and operations teams to clean and load data. This was a cumbersome, slow process and it made it difficult for both seasoned data scientists and general business users to learn from their data,” said Todd Mostak, CTO and Co-Founder, HEAVY.AI. “HeavyIQ breaks down all those barriers, eliminating the need to manually build SQL queries and automatically generating rich visualizations. This allows organizations – including line of business users – to extract insights from their data and make better business decisions.”
With HeavyIQ, HEAVY.AI has taken an open-source LLM foundation model and extensively trained it to excel at core analytics tasks, including analysing massive geospatial and temporal data sets. The technology employs the power of an LLM in conjunction with retrieval augmented generation (RAG) capabilities to take a user’s text input, automatically convert it into a SQL query, and can both visualize and return natural language summaries of results.
HeavyIQ provides many key benefits:
- Accuracy: Trained with over 60,000 custom training pairs, HeavyIQ delivers ground-breaking accuracy. Benchmarks show HeavyIQ to be more accurate than GPT-4, with 90%+ accuracy on common text-to-SQL benchmarks, compared to 85% with GPT-4.
- Speed: Organizations can ask sophisticated questions against large, complex datasets and generate answers (and visualizations) in seconds. HeavyIQ leverages optimized and fine-tuned smaller models that take advantage of the latest NVIDIA GPU hardware innovations to deliver responses up to 10x faster than GPT-4. This allows for quicker iterations and speed-of-thought interaction with data.
- Verifiability: Users can see the generated SQL and verify for accuracy.
- Privacy & Security: A fully locally-hostable model means HeavyIQ can run in customers’ environment and even offline, guaranteeing complete control and sovereignty over sensitive data.
Many organizations are concerned about sharing their data with cloud-based, third-party LLM solutions – particularly government agencies and organizations in highly regulated industries that frequently handle sensitive data. Customers can run HeavyIQ on dedicated hardware located in private, on-prem environments and behind their own corporate firewalls. Data never has to leave their networks.
“We are excited to explore the potential of HeavyIQ as a new query tool for the Harvard Center for Geographic Analysis Geotweet Archive, a global collection of 10 billion Geotweets spanning time, geography, and language,” said Wendy Guan, Executive Director of the Harvard University CGA. “HeavyIQ’s natural language-to-SQL capabilities look promising in providing a more intuitive and convenient way for exploring our extensive Geotweet collection, making it more accessible to a broader range of users with different backgrounds and needs.”
