The BenchSci platform is constantly being updated with new data sources, including our most recent addition of pre-prints. Training our platform to read biomedical research like a Ph.D. scientist would not be possible without our machine learning team. 

We’ve grown a lot during my three years at BenchSci and are only growing more as the company scales up. In fact, we’re actively hiring for dozens of engineering roles right now. 

To give you a peek into what it’s like working in engineering for BenchSci, we’ve asked three of our team members to share more about their experience so far. Ahmed Doghri from our machine learning team and Ramtin Rassoli and Alireza Darbehani from our recently formed machine learning operations teams tell us more about themselves, their work on the team, and how it helps run the BenchSci platform below. 

Watch this video or read on for more of their thoughts on engineering life at BenchSci.


What was your career journey to BenchSci?

Ramtin: I’ve always worked at startups. I’ve always liked the startup culture and life, but sometimes startups don’t have the clearest vision or the vision is not simple or practical. That was not the case for BenchSci. A friend of mine used to work here and when I talked to him about the vision and the product it was both inspiring and practical and feasible. I also joined during COVID so the vision resonated with me. I didn’t have a difficult time choosing BenchSci as my next career move.

Ahmed: I studied computer engineering for my Bachelor’s in South Korea. Later on, I moved to Canada to go to McMaster to finish my Master’s in robotics and AI. There I got introduced to machine learning and all that fun stuff. After school, I joined Scotiabank to work as a product owner for two months. Then I joined IBM to work as a tech consultant in the data science space and there I got introduced to natural language processing which really drove me into the problem that BenchSci is trying to solve. And there you have it, after two years I ended up joining BenchSci.

What skills and languages do you need to know to work in engineering at BenchSci?

Alireza: Teamwork, communication, and technical skills such as Python. Machine learning frameworks and machine learning operations frameworks are a plus for a candidate. Also, general software engineering skills are really important too because when I joined I was surprised. Machine learning technologies at BenchSci versus other companies. The industry is quite mature despite being a young company, we are seriously leveraging the technology.   

Ramtin: Machine learning operations (or ML Ops) at BenchSci sits at the intersection of dev ops, machine learning, and data engineering, so a basic understanding of distributed systems, cloud computing, dev-ops practices, also some fundamental knowledge of machine learning algorithms could help. We write in Python and our infrastructure is on the Google Cloud platform so we use services like BigQuery, Dataflow, and Vertex AI on a regular basis.

What are the biggest misconceptions about your job?

Ahmed: There are two biggest misconceptions I would say! Machine learning is like wizardry, it’s super tough, and you need a Ph.D. to do it. That’s not necessarily true. You just need to have a curious mindset and the ability to peel an onion step by step because a lot of the problems you are going to face, you’re going to treat them as a black box. The other one, especially here at BenchSci, we don’t actually need to know a lot of biology. Having a curious mindset to do the job is of course helpful, but you don’t necessarily need to be an expert in biology.

Alireza: Even for very complicated algorithms there is no magic and there is a way to explain it. Sometimes we just don’t know how to explain it yet. But that doesn’t mean that it’s magic and we shouldn’t trust it.

What's an interesting project you've gotten to work on so far?

Ramtin: One of our main goals in ML Ops is to improve the machine learning engineers’ productivity. One of the grand projects we’ve been involved with for the past year or so was building our unified machine learning operations platform on top of Google Cloud that encompasses all aspects of our machine learning lifecycle from training and inference to monitoring and observability.

Ahmed: We got to use a transformer-based solution which is basically the most state-of-the-art solution out there for natural language processing problems. Basically, we use this model as a classifier for us to actually say if we take two bioentities like a protein and a gene, what is the relationship between those two bioentities within a particular sentence. Is it “influences”? is it “affects”? Is it no relationship at all? We don’t know, we train the model to do that. It reads the sentences and tells you, “Hey, if you take these two bioentities it belongs to this particular class which is a particular relationship.'' The interesting thing about this problem is that the data we were working on was actually highly skewed, it was a highly imbalanced dataset but nevertheless, we were able to achieve really high performances like a 95% +F1 score. It’s actually a super good performance. The accuracy is so high it is actually a practical implementation of this problem. 

What are three words you would use to describe our culture?

Alireza: The first and very important one, which I see the leadership also puts emphasis on, is being open-minded and diversified. Meaning all three letters of D-E-I. Another word I would say is passion. People across the entire team–the technical side, the business side–everyone is very passionate and motivated to make an impact in the preclinical biomedical research space.

Ramtin: Transparent, for sure. Flexible and autonomous.

What are the advantages of working for a remote-first org for you?

Ahmed: I think being in a remote-first company if digital nomad is not the thing for you, even if you’re settling somewhere it gives you an extra edge where you can be farther away from the office if not completely far away from the office. You can stay flexible in terms of where you are in terms of location. 

Alireza: One perk when it’s remote-first is you can work from home or anywhere else in the city. We also have the perk of working from a different city or a different country for a certain amount of time, which is very unique and interesting. You can travel while working and none of the others would be affected. The quality of travel or the quality of work you do.

What is your favorite snack, favorite song, or favorite show?

Ramtin: My favorite song is Green Grass by Tom Waits 

Alireza: I like cheesecake, the trilogy of the Lord of the Rings, and also, Star Wars.

Ahmed: My favorite snack is actually popcorn that I dip into a spicy sauce. It’s a spicy popcorn, you have to try it out!


Our machine learning team is hiring! If this sounds like the kind of team you want to work with check out our open job postings. You can also subscribe to our blog to stay up to date with all things BenchSci.

Written By:
Jason Tang (he/him)