Kaustubh Lall

MACHiNE LEARNiNG ENGiNEER · CLOUD ARCHiTECT · SOFTWARE ENGiNEER · NiCKNAME: KAUS (/KɔS/)

858-***-**** ad2yg3@r.postjobfree.com KaustubhLall kaustubh-lall Education

Master’s of Science in Data Science and Machine Learning 2020 2021 UNiVERSiTY OF CALiFORNiA, SANDiEGO JACOBS SCHOOL OF ENGiNEERiNG San Diego, CA Bachelor’s of Science Degree in Computer Engineering 2016 2020 UNiVERSiTY OF CALiFORNiA, SANDiEGO JACOBS SCHOOL OF ENGiNEERiNG San Diego, CA Published Work

Co author, ”Unique metabolite preferences of the drug transporters OAT1 and OAT3 analyzed by machine learning,” which was published in the Journal of Biological Chemistry (JBC) and PubMed PMID: 318*****-****

Work Experience

Radicle Science Jun. 2021 Present

FOUNDiNG MACHiNE LEARNiNG & CLOUD ENGiNEER Remote, USA

• Owned and developed a reporting system to drive core business needs and implement direct to consumer clinical trials by delivering person alized reports to 50,000 consumers by performing ETL, processing and distributing data using AWS and email notifications. Automated this system on AWS, saving about 40 hours a week, removing the manual bottleneck on scaling the business to handle more studies a month. Skills: Python, Django, Elixir, AWS, SQL, React.js, GraphQL, SES, SNS, Lambda.

• Automated manual data normalization by automating mapping questions across studies into a consistent data warehouse, created and main tained a tool using natural language techniques and data pipeline, expediting analysis across studies and saving 1600 hours across 40 studies. Skills: ETL, Langchain, Elixir, Natural Language Processing, Vector Databases, ETL, GPT 3.

• Partnered with cross functional teams and developed critical infrastructure, managed projects end to end and presented result to founders and senior stakeholders:

Owned and maintained a reminder system for survey completions, helping drive up response rate by improve retention from studies by 10%.

Collaborated on a product recommender system to deliver personalized product suggestions, enhancing the overall user experience and driving engagement by increasing informed decision making in product selection.

Built and maintained an application for real time clinical trial at SupplySide West and other events, driving engagement for 8,000 participants.

Analyzed recruitment and consumer retention data, driving strategic business decisions that reduced recruitment costs by approximately 15%. Collaborated with research teams to publish key findings, contributing to organizational knowledge and industry insights.

Led a team of 3 to develop an internal portal, optimizing and modernizing core business operations, automating data collection and ETL. Skills: Python, Statistics, Scikit Learn, Neural Networks, PyTorch, Django, Docker, GCP. Thorton Hospital, University of California, San Diego Jan. 2019 Jun. 2021 DEEP LEARNiNG RESEARCHER San Diego, CA, USA

• Conducted research on image segmentation, owning a computer vision project to implement a state of the art U Net model to outline bone and cartilage tissue in MRI data helping gauge, and develop quantifiable features for the improvement of osteoarthritis treatments. Skills: Python, Tensorflow, Keras, Convolutional Neural Networks (CNNs), U Net.

• Optimized performance of the image segmentation model and navigated the challenges of a limited training dataset using a combination of interleaving learning and generative techniques, implemented and tuned a model boasting an 0.95+ DICE accuracy on validation datasets . Skills: Data Augmentation, Generative AI, Deep Learning, PyTorch, Image Processing.

• Containerized and deployed the optimized model, accounted for fault tolerance, scheduling and optimized turnaround time. Skills: Model Deployment, Docker, Django, Concurrent Processing. Leichtag Biomedical Research Center, University of California, San Diego May 2018 May 2020 MACHiNE LEARNiNG RESEARCH ASSiSTANT San Diego, CA, USA

• Partneredwithseniorexpertstochosefeaturesanddatasetforanalyticalpredictionofdrug transporterbinding,anddevelopedamodelachiev ing 90%+ classification accuracy on a dataset of chemical descriptors (cheminformatics) related to OAT1 and OAT3 transporters. Skills: Scikit Learn, Python, Statistical Learning.

• Performed rigorous feature analysis and selection to elucidate and explain the model’s rationale and facilitated a logical approach to drug design and explained drug elimination mechanisms, reducing problem complexity from 8̃0 to 5 features. Skills: Feature Analysis, Feature Selection, Dimensionality Analysis.

• Analyzed the underlyingmodeltocollaborateandexplainthemechanismsofprediction,so they can be used for informed drug design. Created analytically tractable solutions, and explainable AI based on the features and decisions made by our models. Skills: Statistical Analysis, Publication.

• Co authored a paper, which got published notably as some of the first research which could explain chemical mechanisms of drug binding and elimination, saving thousands of hours of wet lab testing, and help set cheminformatic ground for studying drug interactions.