Staff Software Engineer, Grafana Labs
Soooo vector databases are pretty popular right now for document search, but it has some drawbacks: choosing the right embedding models, finding the right chunk size, indexing into high dimensional embeddings and managing specialized infrastructure.
If you’re deploying Language/Vision models today, you know that GPUs are the most critical, and often most expensive, component in our infrastructure. Yet, when something goes wrong with a high-stakes training or inference job, the debugging loop turns into slow guesswork.
I get the “why do you live in Seattle” often enough that I decided to write something about hiking. Here’s why I do it:
The Southern California Linux Expo (SCaLE) 21x was held on 15-17th March 2024 in Pasadena, LA. I was presenting a talk titled “So you want to build an Incident Response stack using OpenTelemetry?”.
3+ years, 1200 wonderful new colleagues, $510 Million in funding and 6 acquisitions in, this blog post is a retrospective on what it is like to thrive in a hyper-growth startup. Despite personally having qualms calling a company a rocketship (tweet) I do believe the term applies well to Grafana Labs due to its growing headcount, bullish investments in the observability space, and accelerated adoption from open source enthusiasts to established enterprises.
In Pt.1, I explained that kind of growth I was lucky to witness in my first few years at Grafana. In this post, I’d love to highlight my observations on things that worked well for us. There’s a few things in here that we can do as an IC (Individual Contributor), and a few at the bottom of this list that can be implemented by company leadership. So lets’ get into it.
Pull requests (PRs) are the fundamental building block in any tech company. Every day we create, review and merge hundreds of PRs. Some add features, some fix bugs, some introduce bugs and some are so large nobody wants to review them. However, each of these plays a crucial part in pushing the product and company forward. I previously wrote about best practices for creating a PR here, check that out if you are interested.
So, this is the official documentation on profiles: https://docs.docker.com/compose/compose-file/15-profiles/. Let’s see how we can run multiple profiles with overrides using docker-compose.
From my (limited) experience of working as a software developer in a SaaS company, I believe the most effective way to make an impact on a company and grow your professional career is to stay close to the cost centers. This can mean one of few things -
I joined Cornell Tech as a masters student in Computer Science in August 2022. Here’s my (late) tweet about this new adventure. This blog is a Spotify Wrapped :tm: version of my semester.
This is a blog version of the talk I gave at GrafanaCON 2022.
Contributing to open source is a rewarding activity, but it can be overwhelming to begin with. So I decided to compile a list of gotcha’s that would’ve been useful for me when I got started in this journey, and hope it benefits others as well.
Exciting things are happening, as, for the next few months, I will be working on Enabling Polyhedral compilation in TensorFlow as part of the LLVM community in Google Summer of Code - the annual program by Google to promote Open Source contributions by remote developers.
With AI becoming the latest buzzword in the Computer Science community, it is becoming important, in all areas of computer science, to incorporate concepts from Machine Learning into our workspace. New use cases are emerging for applications of Aritificial Intelligence in biomedical industries, pedestrian detection systems, synthesizing music, etc. How do we construct a neural network that performs a new task?