← Selected Work
Global Management Consulting FirmConsultingBuild · Embed2023 — present

The AI backbone of a top consulting firm's practice.

A board-level AI mandate, a wall of stalled pilots, and one hard question: who actually builds this? We did — 15+ GenAI products now central to how their teams work.

GenAI workbench · live in production
Duration
18 months, ongoing
Team
Pod of 6
Core stack
Next.js · FastAPI · Postgres
Models
GPT-4-class + RAG
Data
~12M documents · 4 TB
Reliability
99.9% · p95 < 2.5s
The challenge

A global consulting firm had a board-level AI mandate and a wall of pilots — none of them in production. Multiple vendors had failed to deliver a prototype that survived contact with real client data and real partners.

The strategy wasn't the problem. Shipping it was.

Our approach

We embedded a senior pod directly inside the firm's teams and started from a partner's actual workflow, not a spec. We built the first tool that earned a partner's trust, then scaled the pattern — every product auditable, access-controlled, and built to the firm's own security bar.

What we built

A connected suite of GenAI products on one platform — research synthesis, drafting, knowledge retrieval, and client-deliverable workbenches — sharing a single design system, SSO/RBAC, audit logging, and an automated evaluation harness.

Tech stack
Application
TypeScriptNext.jsReactNode
AI / data
PythonFastAPILangGraphPostgreSQL + pgvectorRedis
Models
GPT-4-class (Azure OpenAI)text-embedding-3-largeHybrid RAGRe-ranking
Infra
AWS ECS · S3TerraformGitHub ActionsDatadog
What we delivered
  • 15+ production GenAI products on a shared platform
  • Single sign-on, role-based access, and full audit logging
  • An automated evaluation harness for every model change
  • A reusable component & design system
  • Runbooks, on-call, and a knowledge-transfer programme
Timeline
3 weeks
Discovery — workflow mapping with partners & risk
6 weeks
First tool in partners’ hands, in production
6 months
Platform, governance & design system established
Ongoing
Scaled to 15+ products across the practice
Data at work
12M
Internal documents ingested & indexed
80M
Vector chunks in hybrid search
3,000+
Weekly active practitioners
Outcomes
15+
GenAI products live in production
Wk 2
First tool in partners’ hands
99.9%
Platform uptime
Next case →Negotiation intelligence at scale.

Have something
that needs to ship?

Tell us the problem. We'll tell you honestly what makes sense — even if that's not us.

Start a conversation