Skip to content
Est. 2016·Day 0000·
Available for new engagements · Q2 2026
Cairo ·GMT+2·

Mahmoud Amr
Architecture · Data · Cloud & AI · Top-Rated since 2016


§ AI Automation · US fintech (anonymized)

Autonomous AI Sales Agents

Production AI agents that call and text leads from the data warehouse and book appointments 24/7.

Live
in production
<5s
response time
97%
test coverage
Role
Engineer
Client
US fintech (anonymized)
Year
2026
Lane
AI Automation

The problem

A sales org was leaking leads: a fragile no-code SMS funnel dropped conversations, and outbound calling depended on human availability. They needed follow-up that never sleeps and never forgets context.

What I built

Two production agents sharing one warehouse. A voice engine pulls leads from Snowflake and runs a five-phase call cadence with a rotating, health-scored phone-number pool (anti-spam) and timezone-aware dialing. A GPT-4 SMS agent holds multi-turn conversations, qualifies leads, sends context-aware booking links, and auto-follows-up for three days — all serverless, with full observability.

Outcome

Both agents run live in production — sub-5-second responses, 97% test coverage — qualifying and booking leads around the clock with no human in the loop.

Stack
VapiOpenAI GPT-4Node 22Express 5AWS LambdaDynamoDBSnowflakeSQSEventBridgeLangfuse

Client identity anonymized under NDA. Numbers and architecture are real; only names and data are withheld.

Available — Book a call