Work
What the team has shipped
Each of our consultants has 15+ average years of experience driving business outcomes in industry.
Generative AI and LLM engineering
Agent frameworks and LLM applications, built for compliance, conversation and research.
Compliance and security-audit agents
An LLM agent framework for HIPAA-compliance and security-audit automation, fine-tuned with reinforcement learning and served with vLLM.
Conversational agent with retrieval search
A production conversational agent on LangChain.js, with retrieval-augmented search, multi-user chat and federated sign-in.
Multi-agent scoring for deep research
A multi-agent scoring framework behind a deep-research pipeline that quantifies technology-adoption ROI.
Applied research, quantum and high-performance computing
Research programs, published work and the computing platforms researchers run on.
Quantum research at Carnegie Mellon's SEI
Read Jason Larkin's researchSeven years as a Senior Research Scientist at Carnegie Mellon's Software Engineering Institute, as principal investigator of a Department of Defense quantum research program running twenty-one researchers.
DARPA programs and verified control
Work across five DARPA programs, including formally verified control kernels deployed on a robot and on an autonomous car with GPS-spoofing attack detection.
Research computing at the Pittsburgh Supercomputing Center
A production educational high-performance computing platform at the Pittsburgh Supercomputing Center: notebooks, APIs, containers, cluster batch scheduling and observability.
Peer-reviewed quantum publications
Twenty-seven peer-reviewed publications spanning QAOA, VQE, surface codes, magic-state distillation and Trotterization for molecular Hamiltonians.
Regulated-industry and platform-scale machine learning
Machine learning for trust and safety, financial risk and access control.
Trust and safety at X/Twitter and Meta
Seven years building machine-learning detection for spam, coordinated inauthentic behavior and harassment at X/Twitter, then a year on the same class of problem at Meta, serving tens of millions of users.
Risk management at Goldman Sachs
Quantitative research and risk management at Goldman Sachs, building the fund's first formal risk framework.
Access rightsizing at Microsoft
A rightsizing engine at Microsoft using linear programming over historical access activity, with anomaly detection on access patterns.
Data platform, MLOps and observability
Telemetry, observability and model operations underneath production machine learning.
GPU infrastructure observability
A GPU-infrastructure observability platform built end to end: reference architecture, metric set and API schemas across heterogeneous sources.
Telemetry across a GPU fleet
Telemetry ingestion from more than a thousand compute nodes and eight thousand GPUs.
Model operations for commercial lending
Deployment, execution and monitoring for every production machine-learning model at a commercial lending platform.
Leading teams and building capability
Building data science teams, and teaching the people in them.
End-game modeling at Microsoft
Leading end-game modeling for an identity governance platform at Microsoft, after being the first data scientist at three venture-backed startups, all of which were acquired.
Growing a data science team
Growing a data-science team to twelve scientists and promoting its first front-line managers, with a specialty in teaching data scientists to write production code.
Authoring and teaching
Eleven continuous years as a Pearson author across generative AI, deep learning and distributed computing, plus a graduate course taught at New York University.
Enterprise data science training
Designing and delivering data-science training for a Fortune 100 company, onboarding fifteen new data scientists onto an enterprise platform curriculum.
Services leadership
Who leads each part of the work: data science, engineering and infrastructure. Full bios on the Team page →
What a client said about a recent GPU observability build
“We pointed them at some of our hardest problems, and they consistently strategized, self-directed, engaged stakeholders, and executed complex software deliverables at scale. … Early on, we put them in charge of building our observability and monitoring layer to address one of our core problems: SLA enforcement. The work their "full stack data scientists" delivered was excellent! It directly enabled us to secure an exclusive partnership with an industry giant …
They let us focus on our core business while delivering provable short-term wins that justified continued investment in their work. … Would highly recommend!”
On work delivered by Transmute Data, whose consultants deliver BridgeTek's data science engagements. Used with the endorser's written permission; client name withheld.




















