Karan Agarwal

Developer, Guitarist, Geek

[Karan Agarwal @karanagar17]

If you have experience building AI agents at scale, love working in Node.js, and are a fan of Cursor, please reach out.

[Karan Agarwal @karanagar17]

🚀 Big day today — we just wrapped up our roadmap call for Conversations AI at HighLevel, and the numbers speak for themselves:

In March alone -
📅 35K+ appointments booked via AI agents
💬 2.3M messages handled
🧠 1B+ LLM tokens processed

[Karan Agarwal @karanagar17]

🤖 70K+ total AI agents created by our customers

And yet — we're still only responding to ~2% of all inbound messages on HighLevel. Our goal? 80%.

We’re just getting started.

[Karan Agarwal @karanagar17]

(2/n)

- Enhanced scalability: We've scaled to over 30 billion documents without any manual intervention, handling workloads with spikes of up to 250,000 RPS and five million real-time queries.

Link to the full blog -

https://t.co/CttlQVKnJS

[Karan Agarwal @karanagar17]

(1/n) At HighLevel, Firestore serves as our primary application database, and here's what we love about it:

- Increased developer productivity: Firestore's simplicity has boosted our developer productivity by 55%, allowing us to focus on product innovation.

[Karan Agarwal @karanagar17]

- Observability, Alerting, Load Balancers, Istio Mesh, Networking, VPN, etc

If you know someone who has experience with a similar scale, please suggest!

https://t.co/xCLXxY4k5F

[Karan Agarwal @karanagar17]

Hiring a Senior Engineering Manager for Platform team, that manages -

- 200TB+ of application data across Firestore, Mongo, ElasticSearch, Redis, ClickHouse
- 4PB+ of object storage in GCS
- 120Bn events monthly PubSub queues

[Karan Agarwal @karanagar17]

- Rate limiting, AuthZ, and AuthN for 60Bn API hits/month in Node.js
- CI system for 100+ deployments daily for a 300 people engineering team
- 10+ Clusters in Kubernetes with 15,000+ cores of CPU

[Karan Agarwal @karanagar17]

Engineers grow due to two reasons -

Their ability to prioritize what provides value to the users
Solving complex problems in record time