Hi there!
Jalansh is a full-stack engineer and loves to work on different product ideas along with solving high-impact and critical software engineering problems. He believes in the Keep it simple, stupid! principle and focusses on building something quickly and then optimizing on top of it based on priorities. Presently, he works on Adobe Firefly image models and likes to dive into all the stages of model stages. He started off with building user interfaces, but gradually took on responsibilites similar to that of a Machine Learning Engineer. So far, he has dived into data generation, training infrastructure, scientific evaluation, backend services, and rapid prototyping.
Prior to joining Adobe, he worked as a full-stack software engineer on a variety of products in full-time and part-time capacity. Some of the experiences that he cherishes are at The Luminosity Lab, Pirimid Fintech, and Habitica, an open source project for a habit building app. During these experiences, he has worked on a variety of applications and technology stacks.
Jalansh completed his Master’s in Computer Science from Arizona State University (ASU) and his bachelors in Information and Communication Technology from DA-IICT, India. When he is not working, he likes to cook, read, work out, play tennis, go for hikes, and travel. More recently, he is trying to learn a musical instrument. Jalansh is always excited to meet new people in the Bay or online; so if you’re reading this, feel free to reach out!
- Full-stack Engineering
- Artificial Intelligence, Machine Learning
- Generative AI
- Quantum Computing
-
Master of Science, Computer Science, 2022
Arizona State University, USA
-
Bachelor of Technology, Information and Communication Technology, 2016
Dhirubhai Ambani Institute of Information and Communication Technology (DA-IICT) - Gandhinagar, India
Experience
- Making the entire model training simpler one PR at a time by improving the training infrastructure and reducing the time to start training and evaluating a model checkpoint. I have actively been involved in development and PR reviews for changes that are used for production trainings where it is important to continuously determine the quality of image generations.
- Created a new model pipeline in the production repo that brings identity preservation into Firefly’s image offerings. This was an entirely new experience as the parity between research and production pipelines is often not same in the beginning.
- Synthetic data generation, subject to data laws, for the next generation of Firefly Image model. This entails experimenting with a lot of models, to see how the generated data impacts the model because having better data almost always leads to a bump in quality.
- Deploying various models for rapid internal prototyping to see how good the models are, or how can they be used by end users, by using them on a user interface.
- Developed an Android app for a company that aims to reduce stress through wearables.
- Improved the credit rewards system to increase the mobile app usage for Bank of the West.
- Simulating the state of Arizona for analysing the effects of business decisions based on past data.
- Connected small businesses with student volunteers by developing a symbiotic web application.
- UI developer for a web application that lets users post and explore business ideas.
- Revamped the entire prompt UI for providing better customization options to the user for generating images.
- Added seed functionality so that users can get different images for the same prompts.
- Integrated image search for generating creative variations of the searched image.
- Eliminated the need for in-person identity verification between banks and individual users by developing an e-KYC platform. Also optimized the code to reduce multiple API calls to a single API call.
- Migrated stock positions data dashboards to Grafana for improving the efficiency of data filtering.
- Fixed the re-casting of 2 major skills, Stealth and Chilling Frost, which impacted all the users.
- Fixed redundant invites and member count bugs for parties, and helped new contributors to set up the code base.
- Worked on the frontend and backend changes for a 24x7 live platform which receives a high volume of transactions (>100,000) operated by a large US-based MNC.
