Skip to main content

Moin Dalal

Agentic AI Engineer

I build AI agents that hold up in production.

Multi-agent systems, tool gateways, and the tracing and tests that show what every agent did, step by step.

India (IST)Contract or full-timeMBA, Analytics and Strategy

Email me See my engineering workLinkedIn
~/traces

$ replay run-1153 --production

  1. 0.10 sTrigger: bymoin.com contact0.00 s
  2. 0.10 sNormalize: Merge into webhook_input.json0.03 s
  3. 0.13 sQualify: Build prompt0.02 s
  4. 0.15 sKey: scoring pays?0.01 s
  5. 0.17 sQualify: AI lead scoring2.50 s
  6. 2.66 sQualify: Parse AI output0.03 s · tier:HOT
  7. 2.70 sGuard: Check reply0.03 s · guard:pass
  8. 2.73 sConfig: Fetch settings0.20 s
  9. 2.93 sConfig: Resolve0.04 s
  10. 2.97 sAlert: Build Telegram draft0.03 s
  11. 2.99 sAction: Make links0.02 s
  12. 3.02 sAlert: Telegram to Moin0.74 s
  13. 3.76 sAction: WhatsApp links0.04 s
  14. 3.79 sAction: Slack links0.03 s
  15. 3.82 sCRM: Ingest (lead_ingest)1.96 s
  16. 5.78 sRespond: intake result0.01 s
  17. 5.79 sWA: Build alert0.04 s
  18. 5.83 sWA: Alert Moin0.29 s
  19. 6.11 sVisitor: Build ack prompt0.02 s
  20. 6.14 sVisitor: Use AI?0.00 s
  21. 6.14 sKey: writer pays?0.00 s
  22. 6.14 sVisitor: AI acknowledgement2.26 s
  23. 8.41 sVisitor: Build checker prompt0.04 s
  24. 8.45 sVisitor: Check needed?0.00 s
  25. 8.45 sKey: checker pays?0.00 s
  26. 8.45 sVisitor: AI checker1.36 s
  27. 9.82 sVisitor: Build instant email0.05 s
  28. 9.87 sVisitor: Send instant email3.22 s
  29. 13.09 sVisitor: Build send-failed alert0.02 s
  30. 13.10 sRun: Build record0.05 s · outcome:ai_sent
  31. 13.15 sRun: Post log0.35 s
  32. 13.50 sTimer: Wait for the operator0.00 s

0.17 sQualify: AI lead scoring2.50 s · AI call

AI callCode step✓ 32 steps · 13.5 s · outcome ai_sent

13.5 sA full production run of my lead agent: 32 steps, 3 AI calls.
844Automated tests run before every deploy of this site.
9Agents and platforms built, each with its own case study.

Selected systems

Three systems I designed and built.

Each one has a case study with the architecture, the trade-offs and what I would change.

AGENT-07

AI BI Agent

A 4-agent LangGraph system that turns plain-English questions into checked, sourced answers.

LangGraph · Python · SQL

Read the case study
AGENT-08

MCP Gateway

One gateway for every company tool: last-moment logins, rate limits, caching and a tool registry.

MCP · FastAPI · Redis

Read the case study
AGENT-09

AI Agent Observability Platform

Tracing, answer-quality checks and alerts that catch agent drift before users do.

Pydantic AI · LangFuse

Read the case study

Skills, with proof

Each skill links to where I used it.

Grouped the way job posts group them. No list of logos.

Used in

Writing

How I think about agents.

Notes on the trade-offs I meet while building.

Contact

Tell me about the role.

I read every message myself. References are available on request.

Email
hello@bymoin.com
See my engineering workLinkedInBook a 20-minute call