Patronus AI
Patronus AI is an AI research lab developing simulation technology for training AI agents, backed by $70M in funding. The 64-person startup has attracted investment from Datadog, Stanford University, and AI industry leaders.
The growth story
Patronus AI grew from 1 to 64 employees between 2023 and 2026, raising $70M along the way, most recently a Series B in June 2026.
About
Patronus AI builds infrastructure to simulate intelligence for AI training, focusing on digital world modeling and evaluation tools. Their research includes benchmarks like FinanceBench and safety testing systems used in AI development.
The company has grown to 64 employees since its 2023 founding, with engineering making up its largest function. Its $70M Series B suggests rapid scaling in a competitive AI research space.
Founders and leadership
Engineering dominates the org chart with 10 reported roles. Leadership includes investors from Datadog and AI labs, though specific executive names aren't provided in the data.
The Patronus AI org chart
Anand Kannappan (Co-Founder and CEO) leads 14 mapped leaders with 5 direct reports. Click +N to expand a team; names link to the person’s public profile.
Departments, and which ones grew
Headcount by function: Engineering 24 (+71% YoY), Operations 6 (+200% YoY), Business Development 5 (0% YoY), Administrative 4 (+33% YoY), Human Resources 4 (+300% YoY), Product Management 3 (+200% YoY).
Talent flow
Recruits come primarily from acaisoft (4 hires) and tech giants like Meta and Amazon (3 each). Alumni have moved to AI startups including LlamaIndex and Vor, the latter founded by a Patronus alum.
Most common previous employers of Patronus AI staff: acaisoft (4), Meta (3), Amazon (3), NG Logic (3), Facebook (2), ComprandoEnGrupo.net (2). Former Patronus AI employees most often moved to LlamaIndex (1), Sauron (1), Cursor (1), Meta (1), M Booth (1), Abridge (1).
Alumni worth knowing
Web traffic
Traffic mix: 37% direct, 45% search, 12% social, 0% paid.
What this company is made of
Most common skills at Patronus AI: Python (Programming Language) (24), JavaScript (16), Machine Learning (15), SQL (15), Leadership (13), Java (12), Analytical Skills (12), C++ (12).
Where the team actually sits
Patronus AI people are concentrated in San Francisco Bay Area (11), California, United States (10), Mazowieckie, Poland (4), United States (4).
Search footprint
Patronus AI attracts about 255 organic search visits a month, traffic worth roughly $2K a month if bought through ads. It gained 95 ranking keywords and lost 53 in the last period.
What employees say
- Compensation
- 4.50 / 5
- Career opportunities
- 5.00 / 5
- Senior management
- 5.00 / 5
- Work-life balance
- 4.50 / 5
Frequently asked
Patronus AI has 64 employees as of September 2026.
Patronus AI has raised $70M in total, most recently a Series B round in June 2026.
Patronus AI was founded in 2023 by Anand Kannappan, Rebecca Qian.
Anand Kannappan (Co-Founder and CEO) sits at the top of Patronus AI's reporting structure, with 5 direct reports.
The most common previous employers of current Patronus AI staff are acaisoft (4 people), Meta (3 people), Amazon (3 people).
Former Patronus AI employees most often move to LlamaIndex (1), Sauron (1), Cursor (1).
Patronus AI attracts about 255 organic search visits a month, worth roughly $2K a month in equivalent ad spend.
The most common skills among Patronus AI staff are Python (Programming Language), JavaScript, Machine Learning, SQL, Leadership.
Patronus AI staff are concentrated in San Francisco Bay Area (11), California, United States (10), Mazowieckie, Poland (4).
Patronus AI has about 24 people in Engineering, up 71% over the past year, as of September 2026.
Patronus AI has about 6 people in Operations, up 200% over the past year, as of September 2026.
Computed from 45 current and 17 former Patronus AI profiles, as of September 2026. Source: Crustdata.
Build this for any company
Every number on this page came from the Crustdata API. Point the MCP at a company and get the same org chart, alumni list and talent flow in one prompt.