12 agents currently in production

We ship agents.
Not demos.

A boutique AI consultancy out of Gurgaon. Twenty years of enterprise IT, eight in AI/ML, and a workshop full of production-grade agents serving startups and SMBs across India and the US.

20+
yrs enterprise IT
8+
yrs AI / ML
12
agents in prod
04
bootcamp cohorts
// manifesto

Most AI projects die in notebooks. Ours don't. We're engineers first — production reliability, observability, evals, and cost ceilings are the work, not afterthoughts. We ship agents that survive contact with real users.

// what_we_build

Four ways we work.

// core_offering

Agent Engineering

Custom production agents — RAG pipelines, tool-using systems, multi-agent orchestration. Built with evals, guardrails, and cost ceilings from day one.

LangGraphOpenAIAnthropicVector DBs
// consulting

AI Strategy

Roadmaps, build-vs-buy, vendor selection, team structure.

// education

Team Training

Bootcamps and workshops for engineering teams adopting agentic AI.

// products

Custom Products

End-to-end AI-native products from spec to deployment.

// selected_work

Agents in the wild.

Loom Logistics

Procurement Agent

Autonomous RFQ-to-PO agent for a mid-market freight company. Negotiates with 40+ carriers.

63%
cost reduction
8m
avg cycle
v2.4.1
shipped
Northwind Health

Clinical Triage Copilot

HIPAA-bounded triage agent embedded in EHR. Handles intake, summarization, and routing.

41%
physician time saved
99.2%
guardrail compliance
Halcyon Capital

Diligence Research Agent

Multi-step agent for deal teams — pulls filings, builds memos, surfaces red flags.

5x
throughput
$0.18
per memo
// agentic_ai_bootcamp

Train your team to ship agents.

Six weeks. Six modules. Live cohort with the founder. Past four cohorts sold out. Cohort 05 opens for waitlist now.

Cohort 05 — Feb 2026 · waitlist open
~/bootcamp/curriculum
  • 01agent_fundamentals.md
  • 02tools_and_function_calling.py
  • 03rag_pipelines.py
  • 04multi_agent_orchestration.ts
  • 05evals_and_guardrails.py
  • 06ship_to_production.sh
// how_we_work

From notebook to production.

01

Discover

Two-week paid discovery. We map the workflow, the data, and the failure modes — then commit to a fixed scope.

02

Prototype

A working agent in production-shape code. Real data, real evals, no toy demos.

03

Harden

Guardrails, observability, fallbacks, cost ceilings. The boring work that decides whether it ships.

04

Ship & Operate

Deploy, monitor, iterate. Optional retainer for ongoing tuning and capability growth.

// founder_note

AI Agent Café is a boutique by design. Fewer clients, deeper work, engineering-led from start to ship.

"After twenty years building enterprise systems, I started this studio because the industry was drowning in AI demos and starving for AI in production."
20+ yrs enterprise IT8+ yrs AI/MLex-Fortune 100 architectopen-source contributorbootcamp instructor
// let's_ship

Got an agent that needs to ship?

We take on a small number of new engagements each quarter. Reach out with the rough shape of the problem and we'll get back within 48 hours.