Third Street Systems
We help life science technology leaders redesign engineering for the agentic era — from operating model and vendor strategy through hands-on product delivery.
Kendall Square · Cambridge, Massachusetts
The last decade's operating model is giving way to something structurally different.
Large internal teams extended by outsourced capacity are beginning to give way to smaller senior teams directing AI agents — with knowledge captured in the system instead of disappearing with headcount.
This is not a tooling upgrade.
It changes team shape, delivery economics, vendor strategy, and where senior judgment belongs.
01
The model that got us here
A large internal team, extended by outsourced capacity when the roadmap outran it. For fifteen years, this was how enterprise software scaled — and it made sense, because delivery capacity was fundamentally constrained by human hands on keyboards.
02
Where it breaks
Every added person increases coordination overhead as well as output. Context fragments across teams, vendors, documents, tickets, and individual memory. In regulated environments, rebuilding that context is expensive.
Most organizations are adding AI tools to this operating model and wondering why the economics haven't changed.
03
The new unit of productivity
Judgment concentrates. Senior engineers who understand the domain architect the work, direct specialized agents, and review the output. Execution becomes massively parallel while accountability stays human.
The organization gets smaller, seniority matters more, and capacity stops scaling linearly with headcount.
04
What makes it hold
None of this survives contact with a regulated enterprise without a foundation underneath: persistent context, trusted evaluation, traceability, and guardrails that hold.
Get that layer right and two things follow: institutional knowledge compounds instead of leaving with people and vendors, and the audit trail a regulated environment demands is produced as a by-product of doing the work.
Built from operating experience, not a slide deck.
200+
Global engineering organization led across software, data, DevOps, architecture and external delivery.
Days, not months
From problem statement to a working prototype running on real data and real constraints — so the argument gets tested before it gets budgeted.
20+ yrs
Across hands-on engineering, architecture, cloud, DevOps and technology leadership.
AI-native
Hands-on use of agentic engineering to design and build production software today.
We help leaders design the transition — then prove it by building.
Design the engineering organization AI actually requires: team shape, seniority, vendor strategy, economics, governance, and the sequence of change — without disrupting delivery along the way.
Build the foundation that makes AI-native delivery safe and repeatable: context, orchestration, evaluation, traceability, guardrails, and the mechanisms that let institutional knowledge compound. Built for regulated delivery: every agent action is traceable, every output is evaluated before it ships, and the code and data stay inside your environment. Validation evidence comes out of the work itself, not from a project bolted on at the end.
We build. Rapid prototypes to test the model against a real business problem, then production systems when the value is proven.
Founded by Jeremy Thompson, an engineering leader who spent 20+ years building software organizations — most recently leading a 200+ person global engineering organization inside Novartis Biomedical Research.
His experience spans software engineering, data engineering, DevOps, and enterprise architecture. In his final two years at Novartis, he helped move the organization from conventional outsourced delivery toward AI-native engineering while remaining hands-on with the technology.
The person who sits in the strategy review is the person who architects and builds the system that proves it out. That is the whole point of the firm, not a stage it is passing through.
I work at the executive table and in the codebase. Usually both in the same week.
If you're rethinking engineering for the agentic era, let's compare notes.
jeremy@thirdstreetsystems.com LinkedIn