Skip to main content

Work

Real builds. Real numbers.

AI agent builds and products from our studio. Client details are anonymized where required under DPDP and NDA obligations. All outcomes and stack details are accurate.

Filter:
VoiceIndia

68% of inbound calls resulted in a booked appointment without staff involvement.

68% of inbound calls converted to booked appointments without any staff involvement. After-hours bookings increased 3x. The front desk team shifted focus to in-clinic patient experience.

Retell AI ($0.07/min, HIPAA-eligible) · Sarvam Bulbul V3 (Hindi TTS) · Claude Sonnet 4.6 · Twilio
View details
SupportIndiaSales

74% of support queries resolved in WhatsApp, in Hindi and English.

74% of conversations resolved without human involvement. Average first response time dropped from 45 minutes to 8 seconds. Support staff now handle escalations only.

AiSensy (WhatsApp BSP) · Claude Sonnet 4.6 · Razorpay MCP · Shiprocket MCP
View details
Sales

9x faster lead response. 3x more qualified demos booked.

Average response time: 12 seconds. Demo bookings increased 3.1x in the first month. Sales reps now only talk to pre-qualified leads, with company research already completed by the agent.

Vercel AI SDK · Claude Sonnet 4.6 · Apollo MCP · HubSpot MCP
View details
SupportIndia

61% ticket deflection. CSAT improved from 3.8 to 4.6 in 6 weeks.

61% of tickets resolved without human involvement in week 1. By week 6 that figure reached 74%. CSAT improved from 3.8 to 4.6. The support team now has capacity for proactive student outreach.

Mastra (TypeScript orchestration) · Qdrant (vector store) · Voyage AI embeddings · Cohere Rerank v4
View details
VerticalIndiaVoice

250+ leads qualified per month. Sales team now talks only to site-ready buyers.

The sales team now receives only site-ready leads. 250+ qualified leads per month with no manual qualification effort. Site visit show-up rate improved from 34% to 61%.

AiSensy (WhatsApp) · Retell (voice) · Claude Sonnet 4.6 · Sarvam Bulbul V3 (Hindi)
View details
Strategize

41% token cost reduction. Eval pass rate climbed from 67% to 89%.

Token costs dropped from $4,200/mo to $2,480/mo in month one. Agent output consistency improved, with eval pass rate rising from 67% to 89%.

Langfuse observability · Claude prompt caching · Cursor .mdc rules rewrite · CLAUDE.md rewrite
View details

Need a build like one of these?

Book a 30-minute call. We will scope it, price it, and walk you through the human-approval checkpoints we would include.