lastmile ai
LastMile AI was founded to make generative AI useful, reliable and accessible by giving builders tools to develop, evaluate, deploy and integrate AI applications. Its later vision was to build a cognitive computer with persistent context, connections to authorized tools and services, and long-running orchestration among humans and AI.
At a Glance
- Software engineers and product teams
- Machine-learning and AI engineering teams
- Enterprise organizations adopting generative AI
- Developers building AI agents and MCP servers
- +1 more
AI Tools by lastmile ai
(1)
mcp-agent
Python MCP Agent Framework
Discussions
No discussions yet
Be the first to start a discussion about lastmile ai
Latest News
Products & Services
A platform for software engineers and product teams to create, personalize, test, evaluate, deploy and integrate generative AI applications using text- and image-generating models from open and closed providers.
Interactive, notebook-like documents for experimenting with and combining text, image and audio model responses into AI-powered applications and workflows.
A batch-execution and orchestration service for running AI templates with data and chaining model outputs, such as transcription followed by translation and synthetic voiceover.
Reusable development setups that can be parameterized, shared with teammates or the community, and used as starting points for AI applications.
Market Position
LastMile AI positioned itself as a full-stack developer platform for taking generative-AI applications from experimentation through evaluation and production, rather than a model provider. In its 2023 funding coverage, CEO Sarmad Qadri identified LlamaIndex and LangChain as competitors; its later mcp-agent work competes in the MCP/agent-framework and agent-runtime ecosystem while differentiating through MCP-native support, composable patterns, durable Temporal execution and cloud deployment.
Leadership
Founders
Sarmad Qadri
Co-founder and CEO; previously part of Meta's product-engineering team, where he and the other founders built tooling for AI model management, experimentation, benchmarking, comparison and monitoring.
Andrew Hoh
Co-founder; became CPO and President at LastMile AI. Previously worked with Qadri and Sonwalkar on Meta's product-engineering team and its machine-learning developer tooling.
Suyog Sonwalkar
Co-founder; served as CTO during LastMile AI's early period. Previously worked on Meta's product-engineering team with Qadri and Hoh, building tools for machine-learning engineers and data scientists.
Executive Team
Sarmad Qadri
Co-founder and CEO
Former Meta product engineer who worked on AI model-management and experimentation tools before co-founding LastMile AI.
Bill Peterson
CTO
Listed by LastMile AI as its CTO on the company About page.
Founding Story
Qadri, Hoh and Sonwalkar started LastMile AI in 2023 after seeing that software developers and product teams were adopting generative AI while available ML tooling was still designed mainly for research scientists. Their initial vision was to democratize generative AI by consolidating experimentation, prompt engineering, evaluation, orchestration, deployment and integration into one developer platform; the founders' earlier Meta tooling inspired the product.
Business Model
Revenue Model
Developer-platform and managed-cloud model: open-source tooling is distributed through GitHub/PyPI, while the platform and mcp-c cloud deployment were intended to support paid enterprise or managed-runtime usage.
Target Markets
- Software engineers and product teams
- Machine-learning and AI engineering teams
- Enterprise organizations adopting generative AI
- Developers building AI agents and MCP servers
- Executives and knowledge workers
- Building and integrating generative-AI features into applications
- Rapid prototyping and prompt/model experimentation
- Evaluation and regression testing of LLM applications
- Multi-step workflows combining language, image, audio and speech models
- Production AI-agent orchestration and durable long-running workflows
- Deploying agents as MCP servers