Harmony Harmony at LisCon 2021, 20 Oct 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: LisCon 2021 feat. Shresth Agrawal of Paraswap and Harmony Founder Stephen Tse

Description

The Harmony team attended LisCon in Lisbon, Portugal October 20-21, 2021. We met with many amazing people including Shresth Agrawal of Paraswap. Shresth and Harmony founder Stephen Tse discuss designing algorithms, data structure, and bridging chains.

Shresth on Twitter: https://mobile.twitter.com/shresth3103
Discover Paraswap: https://mobile.twitter.com/paraswap
Discover Harmony: https://www.harmony.one

Song Attribution:
7V59IIV8EEWF4DRZ

A

It's really great to restrict from paris swap joining us today. Thank you, yeah sure. Can you tell us about your background. Yeah.

B

So um I am myself, I'm a student.

A

B

Studying at jacobs university in raymond, but I work full-time for parasol as a algorithm designer yeah and I love what I do.

A

Yeah yeah, what does it mean to be an algorithm designer.

B

It's a pretty cool job because you know I need to build like the car of the parasol. Basically- and it's like coming up with ideas to like how do the routing will work and all that stuff. So yeah, that's what I do.

A

Yeah, so data structure is also a big part of algorithm design, yeah yeah. So can you tell us about one of the challenges about data structure for paraswap.

B

The routing itself is quite recursive, because you know you have multi routes and you have like multiple exchanges in them, and I mean we were we kind of refined our structure that we had initially, because initially, we only had like a simple swap architecture like only one exchange or multiple exchanges, but not like a multiple route architecture.

B

So we kind of refine those and also like considering gas into account, was a very big thing, because considering gas into account makes like the whole algorithm pretty complex from something which is very uh linear to something which is very discrete and yeah more much more complex. So.

A

Right what will be also like the next few months next six months, that uh would there be like more changes in the algorithm and data structure that makes things uh like what are some of the research they're exploring.

B

Yeah, so there are a lot of things we can improve on our on our algorithm side. At the moment, we aren't very focusing on the algorithm side we're focusing on other sides, but there is, for example, we want to like optimize like optimize on the gas, so our gas estimates are not like. We we can improve there is. There is scope for improving on the gas.

B

There is also scope on the number of hops that we take and stuff like that. um There is already like some parts which is already written, but we just need to like test it and deploy it. You might see those like in next months by.

A

The time like all these layer, two bridges from other chain, would that make it more complicated or would it help you not to worry about the guest fees, so we.

B

Are already crushed like multi-chain now, because we are already in polygon, avalanche vsc, which is super cool, and that makes aggregation even better, because you don't need to pay much gas and it actually makes sense to use aggregators yeah. It makes things I mean all the evm compatibility is quite easy for us to deploy right, but going crush shane I mean we haven't really thought about it much, but that will definitely be much more complicated. I would say so.

A

From the perspective like you are the algorithm person from from a really I hope from many of the academic classic uh understanding right and now you have to play with on the edm machine. Do you find what you learn from school and which you haven't finished yet to be like something like completely different in practice?.

B

um Personally, I am more of a hacker than like an academy person. I like to read research papers right, but I like to hack on it, so everything that I've learned, I mean I learned things before university yeah everything I learned was like more like applied, so I always like whenever I learned some academic stuff, I would usually apply it like try to write some code right and playing with evm. I actually started you know two years ago to three.

A

Years ago so.

B

It's like, I have a bit of experience and it's just fun actually yeah, uh it's a bit limited, but evm can do. But with that limit you can do a lot of things. Absolutely.

A

And actually forces you to understand the trade-offs exactly and the fees make everyone really spend time studying.

B

Their algorithms yeah, like optimizing and gas, is such a big part as an aggregator like we really spend hours just like optimizing that single bit of gas so that you just pay less gas yeah it's painful, but you have to do that.

A

So if I give you even like two three years time frame more than just hacking engineering effort, what kind of open research problem, as you call a cell algorithm designer that will get you excited wow in two years, maybe this big problem will be solved. Oh, like an.

B

Algorithm's problem yeah, um I don't know I've never thought about it like it's like two years down. The line is a big thing.

A

It's like all these, like, like a minor extractable thing right, like yeah, there'll,.

B

A

Quick aggregation level, there's always that between the time tradeoffs and fees trade off and how complex is your data structure. I mean.

B

Yeah, if you think about it, like you know, if you go like crush chain arbitrage, for example, maybe in like six months already, that would be something you can do so.

A

B

You can like do composable transactions cross chain and do arbitrage across them. I mean that would be something very interesting, but I don't know if it's an algorithm's problem, but it's more like a general thing. I think it will be interesting.

A

Especially, maybe there can be a slash board. Flash flash loan should be actually like be part of the whole transactions that you have to.

B

Yeah, you can use flash loans, you can use so many d5 primitives, like the number of d5 primitives. Now we have is increasing day by day yeah. You can take loan, you can deposit here, you can get this yeah derivatives. There is leverage, there is a flash, I don't know yeah and there will be nfps there. You can fractionalize them and then I don't know whatever you can do with them.

B

But it's just crazy and at some point we will have so many primitives like thinking about something like all these rving and everything will become so much more fun. I would say.

A

I can tell your examiner and passion thank you for joining us today. Thank you, it was. It was really nice talking to you my pleasure.