Siva Gurusamy

principal architect at Harvey Ai, previously at @motive_inc, Twitter, ex-googler, and life long learner

[Siva Gurusamy @sivaag]

If you are in a room full of explosives, don’t blame someone if he throws you a match.

[Siva Gurusamy @sivaag]

We’ve built our own model.
Now helping law firms build theirs with their private data.

If sounds interesting, pls dm me or Gabe.

[Gabe Pereyra @gabepereyra]

When I shared @harvey’s model strategy a few months ago, there were two parts:

1. Build our own model
2. Use that to help customers do the same

We’ve done the first. Now we’re hiring for the second: Harvey’s Private Model Program.

We’re seeing huge demand from law firms to own their own intelligence leveraging private data.

This will enable them to become frontier firms that get smarter with every client matter.

You’ll:
 - Partner with law firms to build these systems.
- Build the team that delivers this at scale.
 - Work with our technical org to define the platform that powers this team.

We’re looking for a technical PM or founder type to own Harvey’s Private Model Program.

DM me if this sounds interesting.

[Gabe Pereyra @gabepereyra]

In July, 20% of our inference spend was review tables. The most expensive review table queries cost $20k.

Today we’re excited to share work we did with @appliedcompute that will reduce this cost by 50% by post training a review table model.

Our review table product allows layers to upload thousands of contacts or emails and extract or analyze flexible fields.

Historically we optimized this system via prompt caching, packing (multiple fields per model call), routing, batching, etc.

Once we hit the limits of gradient free approaches we wanted to see if post training could improve on an already strong baseline.

Review tables turned out to be a perfect candidate for post training because verification is easy and we’ve built large datasets to do so.

Awesome work by @vtrengarajan, our vault and labs team as well as applied compute. Detailed write up below.

[Thiago Ghisi @thiagoghisi]

Ten Principles for Growth as an Engineer by Dan Heller:

“These are the most important lessons that I wish I had learned ears earlier than I did; I sure wish someone had sent it to me when I was 22.”

1. Reason about business value: Reason like a CEO. Understand the value of your work to your company, and take responsibility for reasoning about quality, feature richness, and speed. Your job isn't just to write code; your job is to make good decisions and help your company succeed, and that requires understanding what really matters.

2. Unblock yourself: Learn to never, ever accept being blocked; find a way by persuasion, escalation, or technical creativity. Again, your job isn't just to write the code and wait for everything else to fall into place; your job is to figure out how to create value with your efforts.

3. Take initiative: The most common misconception in software is that there are grown-ups out there who are on top of things. Own your team's and company's mission. Don't wait to be told; think about what needs doing and do it or advocate for it. Managers depend on the creativity and intelligence of their engineers, not figuring it all out themselves.

4. Improve your writing: Crisp technical writing eases collaboration and greatly improves your ability to persuade, inform, and teach. Remember who your audience is and what they know, write clearly and concisely, and almost al-wavs include a tl;dr above the fold

5. Own your project management: Understand the dependency graph for your project, ensure key pieces have owners, write good summaries of plans and status, and proactively inform stakcholders of plans and progress. Practice running meetings! All this enables you to take on much bigger projects and is great preparation for leadership.

[Siva Gurusamy @sivaag]

Bald eagle. One of very few in Bay Area.

[Siva Gurusamy @sivaag]

Glad so many great people are working AnythingGPT. And no more on speculative web3/crypto.

Productivity/Knowledge should go to next orbit in next 2/3 years.

[Siva Gurusamy @sivaag]

@marcinkadluczka If you assume unconstrained, B or C. It’s difficult predict initial trend.

[Siva Gurusamy @sivaag]

Glad and happy. Hope water fills up in Shasta lake. Pretty sad to see red lake.

[Colin McCarthy @US_Stormwatch]

California's Sierra Nevada range 80 days ago versus today.

These mountains likely currently hold the biggest snowpack on the planet.

[Siva Gurusamy @sivaag]

@jianhangg Adserver will miss you. Great working together. Wish worked longer with you.

[Siva Gurusamy @sivaag]

@elonmusk Point to ponder: why does anyone go to instagram or TikTok?

Twitter is place for news & breaking events.

[Siva Gurusamy @sivaag]

@elonmusk Yea. Twitter is way better. But, do we really want this kind of event to happen every week?

Twitter has always been super active during negative. Yes, some positive events as well.

This mindset has to change for Twitter to succeed.

[Siva Gurusamy @sivaag]

How often engineers are nitpicking on PR reviews?

They should really be catching logical error.

Or is it a lack of broader context beyond the code being reviewed?

[Siva Gurusamy @sivaag]

Anyways, discovering best content to read is hard - google still don’t solve this problem.

Or we just don’t have enough time to read.

[Siva Gurusamy @sivaag]

Medium and Pocket are under appreciated apps.

Medium has excellent content library ever written but poorly organized.

Pocket is getting commoditized with twitter - read in app. But, for other apps, pocket is way better to read.

[Siva Gurusamy @sivaag]

This must be hardest thing to focus.

Thanks Tweeps. Great way say good bye. 🫡 💙

[Laura Froelich ✈️🏈 @laurafrofro]

This is emblematic #OneTeam

[Siva Gurusamy @sivaag]

Tweeps, be strong. this may shall pass.

Will be happy to chat for open roles, references or just someone to talk.

[Siva Gurusamy @sivaag]

It feels like years of progress in few days in Twitter.

Finally they have a good PM and Salesperson as a Chief Twit?

[Siva Gurusamy @sivaag]

backlog -> blackhole - find me if you can!

[Siva Gurusamy @sivaag]

Identifying what is signal and noise can be a good skill.

Some may think signal is noise. These are alerts configured poorly.

Some may think noise is signal. They are expert in that system/domain.

[Siva Gurusamy @sivaag]

(c) db or data store going down.

Most often big companies have nosql. Companies with relational db down is another cause of outage

[Siva Gurusamy @sivaag]

How often full outage in large companies is due to (a) code bug vs (b) configuration change in tuning some underlying infra (not the code logic).

Most of then it’s configuration change. Why? Untested code :)

DNS config changes can lockout everyone.