At Reverie Labs, we’re building a pharmaceutical company from the ground up using computation—we’re a biotech company that looks and feels like a tech company. We're focused on using machine learning and computational scale to solve challenging problems in cancer drug discovery.
Cloud Architect Boston, MA, United States
DevOps Engineer Boston, MA, United States
Front-End Software Engineer Boston, MA, United States
Full-Stack Software Engineer Boston, MA, United States
Infrastructure Engineer Boston, MA, United States
We're an all-engineer and scientist team, with experts in machine learning, medicinal chemistry, and computational chemistry working together to power therapeutics programs.
We have a wide range of tools and resources in our technical stack, and pride ourselves on giving engineers ownership over this work. This includes machine learning models, data ingestion workflows, front-end applications built in Django and other frameworks, and scalable infrastructure on AWS and GCP.
We're backed by strong investors. We were in the W18 batch of Y-Combinator, and after that raised money from top-tier firms like First Round Capital.
Our engineering team is deeply collaborative and has a rapid-prototyping culture. We regularly read a paper or discuss a new idea, have a design meeting the same day, and have a working prototype within 24 hours in the hands of our chemists to evaluate feedback and make improvements.
We have a weekly all-hands where each person shares a summary of their deliverables in the past week, and gives a roadmap of the upcoming week. The CTO also does weekly 1:1s with each engineer.
All code is checked in into repos in our Github organization, and we have a pull request review process that matches our broader collaborative culture. We also have built substantial CI/CD infrastructure using both CircleCI and Github Actions, which allows us to unit test our code, build development and production containers, and ship those containers automatically upon merging a PR. We do this multiple times per day.
We deal with a wide variety of technical challenges, but here is a summary of a few:
Every day, we get new data from our labs on the outcomes of chemistry and biology experiments. This means that we are getting new training and evaluation data daily for our machine learning models, which need to be correctly labeled and integrated into a central data lake, and then used to retrain a large suite of models. We train models for dozens of chemistry and biology tasks, which are all necessary to make decisions about which compounds to progress to the clinic, and so we have built infrastructure to manage these workflows.
Our tools and applications are all internal-facing, since they are used to power our internal drug discovery programs. As a result, an engineer can work closely with chemists and biologists to understand their needs and build tools that improve their decision-making abilities.
We work with a number of third-party lab service providers that need access to certain subsets of our data, either because they use that data in the lab or because they are the source of the data in the first place. As such, we have architected an AWS environment with multiple virtual private clouds that allow us to securely segregate resources and maintain clean access control settings.
Our engineers generally have not had chemistry backgrounds prior to joining Reverie, and this is not expected! They have been able to quickly integrate a wide variety of chemistry visualization and manipulation packages that allow us to use our computer science abilities to power drug discovery programs.
Full-Stack Engineering: We architected a versioned data registry system that allows users and automated systems to create datasets that are used to train machine learning models. Datasets are be updated regularly, and versions must be static to allow for reproducibility. The system handled the consistent construction of train/valid/test splits. Furthermore, this system tied closely with an internal leaderboard that allows us rapidly evaluate what the best models are for each task and deploy those to our production environment.
Machine Learning, Full-Stack Engineering, Front-End Engineering: Retrosynthesis Planning Interface. Retrosynthesis planning is the chemistry problem of describing how to synthesize a molecule of interest, or determining that it is not synthesizable. Compounds are synthesized through a series of reactions that form intermediate compounds that lead to the final product. As a computer science problem, this is essentially a search problem that you can use methods like BFS and DFS to solve, along with a variety of heuristic-based and reinforcement-learning-based variants. In a recent project, an engineer (with no chemistry background!) learned how to use reaction chemistry to implement a search algorithm that did retrosynthesis planning in a fraction of a second. This engineer worked with a fellow engineer to build a front-end application that allows our chemists to use this tool to determine whether their compound ideas are synthesizable, along with a batch script that lets us evaluate compounds in bulk.
DevOps Engineering: We internally serve a variety of our tools and services, including both machine learning models and web applications built in Django. For all of these, we built a Kubernetes cluster (using AWS's Elastic Kubernetes Service) that gives us a scalable framework for serving predictions on billions of compounds with dozens of machine learning models. For our web applications, we used the external-dns Kubernetes pod to give DNS resolution using AWS Route 53, and this allows our employees to access tools when connected to our AWS Virtual Private Cloud. To further facilitate this connection, we set up AWS Client VPN, which gives a secure and easy-to-use connection to our VPC from any employee's Mac, enabling secure access to all of our tools and services without needing to expose them to the public internet.
We foster a scrappy, rapid-prototyping mindset to our work. Engineers are encouraged to take time to try bold ideas and seek feedback early and often to improve the quality of the outcome. We don't like reinventing the wheel - drug discovery is already hard enough. Engineers are encouraged to use and contribute to open-source software, allowing us to quickly build solutions and focus on advancing the science of drug discovery.
Reverie has an unlimited vacation policy.
We provide free snacks in our office! We also have breakfast essentials, and allow employees to expense Uber Eats deliveries if they are staying late.
We heavily discount access to a local premium gym, VIM Fitness. Employees only pay $10/month (as opposed to $80).
We're located in the heart of the biotech epicenter in Cambridge, MA! We're only a short walk from MIT, Harvard, and the Charles River. We designed our office to be bright and inviting, and encourage employees to work wherever they are comfortable.
A few times a year, we have fun company retreats into the mountains of New England! Last year, we went to New Hampshire and rented a cabin and hiked as a team. We are about to have another retreat in a couple of months.
Most of our engineers are in the office between 10am and 5pm, but it's not a hard rule. We generally have pretty flexible hours and value outcomes as opposed to hours.
Full medical, dental, and vision coverage. 100% of employee premiums paid.
We go out to lunch as a full team every Friday to a Cambridge-area restaurant. We also have monthly game and movie nights, and employees often play board games or Bananagrams in the evening.
We offer a pre-tax commuter benefits for monthly passes on the Boston T.
We use Guideline as our 401(k) provider. Reverie will match 100% of your contributions up to 3% of your base salary, followed by 50% of your contributions up to 5% of your base salary.
We let our engineers work from home when they need to.
Interested in this company?
Skip straight to final-round interviews by applying through Triplebyte.