Given a LinkedIn database, we were tasked to efficiently determine companies and individuals that fit a specific ICP and automatically write queries to pull that data. We built ML models that could write and execute SQL queries from a natural language input, and learned over time from previous queries. The result enabled Blueprint to query their data and complete engagements much faster, driving over $600,000 in revenue in 6 months.

Given a list of mentors and companies, we were tasked to identify which mentors would be most relevant for each company depending on stage, geography, industry, and more. We built a powerful RAG system that could surface the best mentors for each company given a ton of unstructured data about all the companies and mentors. The result enabled our client's 200-person Entrepreneur Experience team to do their jobs significantly faster and better by consistently finding the most relevant mentors.

We were tasked to synthesize information from past conversations, CRM, Airtable, Slack, Google Drive, and more to help with VC due diligence when evaluating companies. We built best-in-class enterprise search: custom ranking functions, term significance, re-rankers, query expansion, contextual embeddings, and more. The result enabled our client to perform due diligence faster and make better decisions by surfacing long-forgotten information that contributed to go/no-go decisions.

We were tasked to automate ad hoc data requests to reduce amount of time data team is spending on redundant requests, and enable non-technical users to access data faster. As a solution, we built an AI agent that learns from previous queries, can write SQL and Python code, reflects, plans, and asks the user questions when needed. All human-in-the-loop.

We were tasked to enrich data using various tools, pull similar information from thousands of files–all in different formats–and prepare this data for analysis. Our solution is a table of AI agents that can use various tools including web search, enterprise search, and more. This build reduced time/cost for lead enrichment and sales prospecting, and enabled analysis of data that was previously impossible to analyze.

Our client needed to understand where AI systems in a big bank are finding their information and ensure that the information is accurate and traceable, not hallucinated. As a solution, we built a source attribution platform where our client could connect all of their data sources, and then determine which documents (if any) specific information came from. As a result, we turbo-charged bank’s AI governance efforts, which are top priority in light of recent legislation.

We were tasked to generate monthly financial reports for each of a venture capital fund’s portfolio companies. We built a system that learns format and tone from previous reports, and generates new reports in a similar format/tone, automatically using new data for the content.