27 | Founder 🇺🇸 AI engineer | MLOps | SaaS Founder | building my own company.
California, USA
Joined November 2026
- Tweets714
- Following781
- Followers986
- Likes408
Perfect time to launch. 😂
Microsoft experiencing outages in outlook and other apps today
Introducing Claude Meetings: add Claude to your team’s video call and it builds docs, whiteboards, and other artifacts with you while you’re still talking.
Starting to roll out in beta on Team and Enterprise plans.
1
823
What’s the point of having billions if you don’t have luxury to live your life:
NVIDIA announces CEO transition. Read the release: nvidianews.nvidia.com/news/n…
22
Secrets tip that actually sticks:
Never put long-lived credentials in environment variables or config files that get checked in.
Short-lived tokens + a secrets manager + automatic rotation reduces the blast radius when something leaks.
1
26
CI/CD reality:
A pipeline that takes 40 minutes to give feedback is almost as bad as no pipeline.
Fast feedback loops (under 10 minutes for the common path) change how people write and review code.
26
Terraform tip:
Use data sources and remote state carefully, but avoid giant monolithic modules.
Small, well-named modules with clear inputs make reviews faster and mistakes harder.
1
48
Building in public note:
The hard part isn't shipping the first version.
It's deciding what to measure so you know whether the second version is actually better.
7
4
77
Small SRE win:
Document the exact steps to roll back a deployment in under two minutes.
Then practice it once a month.
Speed of recovery matters more than never failing.
24
For anyone shipping ML models:
A model that works in a notebook but fails silently in production is worse than no model.
Add simple drift detection and a clear rollback path before you chase accuracy points.
2
2
40
Cloud cost reality check:
The biggest savings rarely come from switching instance types.
They come from deleting unused resources, fixing chatty cross-AZ traffic, and stopping services that no longer have owners.
1
34
Observability habit worth stealing:
Every new alert must include a link to the exact dashboard or query that proves it's real.
Cuts down noise and makes on-call less painful.
22
Question for people running production systems:
When a service starts flapping, what's the first thing you check that most people skip?
(Mine is usually recent config or secret changes.)
1
43
Kubernetes tip that saved us real money:
Set memory and CPU requests properly, then use VPA or Goldilocks to right-size over time.
Most clusters are over-provisioned by 30-50%. The difference shows up on the bill, not just in the dashboard.
1
1
42
If you're early in AI/MLOps:
Stop chasing the newest model every week.
Start measuring latency, cost per request, and failure rate.
Those three numbers will teach you more than any leaderboard.
6
9
225
Question for the builders:
What's one practical thing you automated in your ML pipeline that saved you the most time?
(Not the fancy one — the one that actually worked.)
2
115
From big tech to building my own thing:
The meetings got shorter.
The accountability got longer.
Anyone else feel that shift after leaving Amazon or Walmart?
1
2
125
Simple MLOps reminder:
You don't need a perfect platform on day one.
You need a way to retrain, know when it's wrong, and roll back.
Everything else is optimization.
1
3
74
Honest question:
What's your most expensive AI habit?
(Long agent loops, image gen, rewriting prompts 20 times, etc.)
1
1
71
Quick question:
Which AI is costing you the most tokens right now — Claude, GPT, Gemini, or Grok?
And which one do you still use because the results are worth it?
11
10
308