Boston is a city of universities and laboratories. It is known for biotechnology and pharmaceuticals, for research, for robotics, for hospitals and for asset management. Many of the people who would read this page are scientists or engineers themselves, and they will want to know how a claim is tested before they care what is being sold. So this page is mostly about method.
Biotechnology companies
A young biotechnology company has few people and a great deal to read: papers, protocols, its own results. What it tends to need from AI is modest and exact. It wants to query that material and get answers that cite it, and it wants small tools that remove hand work between instruments, spreadsheets and reports. The services are RAG and data platforms and internal tools with AI.
No case study on this site comes from biotechnology or pharmaceuticals. What transfers is how I test. My research on agents is pre-registered, which means the prediction is on paper before the measurement is taken. The same rule runs through my LLM evaluation work. A judging model is calibrated against human labels, and a verdict is taken by majority across runs, so that one lucky sample decides nothing.
Robotics companies
Here the limit comes first. Nothing on this site is about perception, planning or motion control, and I do not offer them. No case study here comes from robotics either. A company that builds robots still has a great deal of software around the machine: manuals and service notes that technicians search, support questions from customers, records from units in the field kept in several systems. That is where an assistant or an agent earns its place.
What transfers is closer than it sounds. For a lean manufacturing practice that works with plants, I built a documentation corpus, an assistant that answers only from it, and an MCP server that lets an AI agent read the same material a person reads. The case study describes all three.
Universities, hospitals and asset managers
Universities have a direct match on this site: the evaluation gates of a multi-agent system that serves a higher-education group with 700,000+ students. Hospitals do not. No case study here comes from one. The health work on this site is a COVID-19 testing platform for a laboratory, operations software from 2021 and 2022, not an AI system. What transfers to a hospital is the pattern of that education system: answering, scheduling and handing over to a human team. Asset managers ask for a record of what an agent did and with which data. The guide on agent auditability says what that record must include, and I build to it.
Boston time and Monterrey time
Boston is on the Eastern clock it shares with New York: one hour ahead of Monterrey from early November to mid March, and two hours ahead from mid March to early November. Monterrey never changes its hour. The work itself sits in your repositories and cloud accounts, and review is done on the code and in writing.
Starting, and what is not on offer
Send the role or the problem. I read the process first and answer with a first build, the access it requires and what I would not promise. Both senior roles and contract engagements are open. A person in Boston is not on offer, and neither is the science: your researchers own the questions, and I build the system that helps them answer.