News

• August 2025

  Deeply humbled to hear many of LLM Evals expert Shreya Shankar's favorite topics are covered in our most recent paper!

• March 2025

  Check out new my blogpost about Constitutional Classifiers!

Skills

Languages

• Python
• TypeScript
• Java
• Elixir
• Clojure

AI/ML

• GenAI (GPT 4o, Claude Sonnet 4, Cohere Command A)
• Langchain
• HuggingFace
• PyTorch
• LLM Evaluations
• Benchmarking LLMs
• Scikit-Learn

MLOps

• ML Pipelines and ETL
• CI/CD
• AWS (EC2, S3, Terraform)
• Elasticsearch
• Docker

Industry Experience

• AI/ML Engineer; University of California, San Francisco - San Francisco, CA

  September 2023 – Present

  Architected and deployed clinical AI systems to save social workers 300+ hours/ week, identify patients at risk for readmission, and create simple, interpretable models from multimodal patient data

  Drove end-to-end architecture of a GPT 4o powered clinical summarization pilot to condense patient charts (20K+ tokens) into actionable summaries, co-developed with 5+ clinicians to improve patient care

  Established and led human and automated annotation pipelines with active learning to ensure 90%+ accuracy on 1000+ GPT 4o clinical summaries

  Designed and deployed a scalable ETL pipeline to process health record data, enabling ML training and evaluation for 3000+ patient records and 30000+ patient visits

  Engineered, deployed, and evaluated a custom readmission risk prediction model, adopted by 10+ clinics, with 12% higher accuracy than the general model

  Created and presented a hybrid LLM-empirical framework to reconcile clinical LLM hallucinations, achieving 40% interpretability improvement (validated by clinical experts) while matching top predictive performance and enabling production deployment

  Developed and authored a publication for a multimodal ML pipeline to integrate tabular and unstructured clinical data to improve black-box model interpretability, delivering 7% AUC improvement in readmission risk prediction

• ML Engineer; University of California, SD; CA

  June 2022 – Mar 2024

  Developed a model-agnostic bias auditing framework surfacing bias in up to 30% of lending datasets across 5+ models and 3+ datasets with 25+ real-world constraints

  Collaborated with a lending expert to define real-world constraints on lending datasets to identify individuals subject to preclusion

  Authored a publication that received an ICLR Spotlight (Top 5% among submissions)

• Software Engineer; Edovo - Chicago, IL

  Jan 2020 – May 2021

  Led the re-architecture and deployment of a scalable educational platform, handling the full lifecycle from leading product requirement sessions to engineering a data backend

  Architected, tested, and deployed a content platform using Elasticsearch to handle 700K+ requests per day to thousands of users

  Led 10+ requirement gathering sessions with Product owners to re-build a platform

  Created a data pipeline and job to merge 4B rows of user event data in PostgreSQL

• Lead Software Engineer; 8th Light - Chicago, IL

  Aug 2017 – March 2019

  Built high-performance technical systems from inception to production, including load testing platforms, API integrations, and patient-facing healthcare applications

  Implemented and deployed a scalable load testing platform simulating 1000+ RPS

  Deployed API integrations to sync 1000+ interactions/ minute in different timezones

  Developed a diabetes management iOS app to connect 60+ patients with diabetic nurses

  Mentored peers and residents through pair programming sessions and code reviews

Papers

• When the Domain Expert Has No Time and the LLM Developer Has No Clinical Expertise: Real-World Lessons from LLM Co-Design in a Safety-Net Hospital

  Avni Kothari, et al

  under review, 2025

  This paper introduces a co-design framework for LLM applications in under-resourced healthcare settings that (1) decomposes summary generation into discrete attributes, (2) iteratively optimizes each attribute, and (3) refines and validates them through a multi-tier cascading process to ensure accuracy, comprehensiveness, and verifiability.

  Concept Bottleneck Models with LLM Priors

  Jean Feng, Avni Kothari, et al

  under review, 2025

  This work eliminates the need for human-annotated concepts by proposing a novel method to learn concepts by wrapping LLMs within a Bayesian framework. This approach is highly generalizable across various data modalities and allows for rigorous uncertainty quantification despite LLMs being prone to error and hallucinations.

  Prediction without Preclusion: Recourse Verification with Reachable Sets

  Avni Kothari, et al

  ICLR – Top 5% among submissions , 2024

  Individuals can be assigned predictions that they cannot change through actions on their features. This paper investigates and formalizes scenarios of predictions without recourse. We argue the importance of these scenarios for both model development and recourse detection methods.

  Bayesian Priors From Large Language Models Make Clinical Prediction Models More Interpretable

  Avni Kothari, et al

  AMIA - American Medical Informatics Association, Podium Abstract, 2024

  Implementing a Predictive Model to Reduce Hospital Readmissions in a Safety Net Healthcare System

  Arturo Gasga, Avni Kothari, et al

  ML4H - Machine Learning for Health, Oral Spotlight , 2024

Research Software

• bc-llm

  June 2024

  Responsible AI software that enhances model transparency and safety through a multimodal LLM–Bayesian framework, allowing for the creation of simple and interpretable models.

• reachml

  June 2022

  Created a model-agnostic fairness and safety audit to identify scenarios of preclusion

Education

• University of California, San Diego

  Masters in Computer Science

  2021 – 2023

  Thesis: Foundations of Model Agnostic Recourse Verification

• University of Texas at Austin

  Bachelors in Mathematics and Economics

  Minor: Computer Science

  2011 – 2016

Poster Presentations

• NeurIPS Workshop Statistical Foundations of LLMs and Foundation Models

  December 2024

• ICML Workshop on Data-centric Machine Learning Research

  July 2023

• ICML Workshop on Spurious Correlations, Invariance and Stability

  July 2023

• ICML Workshop on Artificial Intelligence & Human Computer Interaction

  July 2023

Teaching Experience

• Fall 2022

  TA for DSC 291 - Interpretability and Explainability in Machine Learning

• May 2011 – August 2013

  Differential Calculus Tutor

Service

• Vision 1948 (May 2023 – Present)
• PenPal for the Incarcerated (September 2020 – Present)
• Warren Community Garden (August 2021 – Present)
• UCSF AI4All (July 2024 – July 2024)
• The Recyclery (August 2018 – May 2021)