Custom AI software and consulting for businesses with valuable data
We build natural-language data interfaces, multimodal analytics tools, and AI-powered data pipelines, and help teams cut the cost of running them.
SELECT category, growth_pct
FROM quarterly_sales
WHERE quarter = 'Q3'
ORDER BY growth_pct DESC LIMIT 4;Answer: outdoor and kitchen grew fastest, up 18% and 14% over the previous quarter.
What we build
Every project starts from a concrete business question. Our tools work on real, messy data and are built to keep AI costs under control.
Let people ask questions in plain English and get answers straight from your databases: text-to-SQL, chatbots for data, and data agents that run multi-step analyses. Designed to be reliable and auditable on your own schema.
For example: A chat interface on your warehouse that answers with a query, a chart, and an explanation.
Ask questions across photos, documents, and tables in one place. We build a working prototype on your own data, so you can judge the value before committing to a full build.
For example: Search property listings by what is in the photos and what is in the description.
Turn unstructured images, PDFs, and product catalogs into clean, structured data with tags and attributes generated at scale. Then build search and matching interfaces on top, designed around how your customers actually look for things.
For example: Extract style and material attributes, then help customers find furniture pieces that go together.
Consulting
LLM-powered SQL, chatbots, agents, and data pipelines can get expensive fast. We review your existing setup, on platforms such as Snowflake Cortex, BigQuery, or Databricks, and recommend concrete changes that reduce cost without sacrificing results.
We trace which queries, prompts, and models drive your AI spend.
Match models, prompts, and execution strategies to each task instead of using the biggest model everywhere.
Savings are measured against a quality bar we agree on up front, using your own data.
How we work
A 20-minute conversation about your data, your goals, and what a useful result would look like. No cost, no obligation.
We agree on a clear scope and timeline, then deliver working software on your data, typically within weeks.
Extend the prototype toward production with us, or take the code and documentation and continue in-house.
About
AI-First Analytics was founded by Immanuel Trummer, Associate Professor of Computer Science at Cornell University. His work sits at the intersection of data analysis and large language models: natural-language interfaces to data, AI agents that analyze it, and efficient processing of text, tables, images, and sound.
He received an NSF CAREER award and several “Best of VLDB” and “Best of SIGMOD” distinctions, and wrote Data Analysis with LLMs, now available in five languages. Clients work with him directly: senior, hands-on engineering, with no layers of account managers.
Learn more about Immanuel TrummerSelected work
Book
Text, Tables, Images and Sound
A practical guide to using large language models to analyze text, tables, images, and sound, published by Manning and available in five languages.
View the book at ManningA 20-minute call is the easiest way to find out whether we can help. Tell us what you are trying to do and we will tell you honestly what is feasible.
