Aakash Mishra

I build medical imaging and language models at Epsilon Health, an AI-native radiology practice in San Francisco. Before that I spent three years at Meta, most recently as a senior machine learning engineer and area lead in Monetization. I studied computer science at Harvard and Stanford.

Aakash smiling at a counter with a mug of hot chocolate

Epsilon Health

Epsilon is an AI-native radiology practice. Our models support Epsilon's own radiologists as they read X-rays. I work on how those models see, explain and report.

A foot radiograph passing through Epsilon's CR preprocessing pipeline
Aug 19, 2026 · Epsilon Health Research

Going Beyond ImageNet Preprocessing for X-ray Vision Models

Building a CR image preprocessor from the ground up
Report-model attention moving across a hand radiograph as each finding is written
Jul 29, 2026 · Epsilon Health Research

Auditing Our Models for Visual Explainability

Is learned attention sufficient for faithful and well-localized heatmaps?

Papers and reports

All 14 papers
System diagram: an upstream causal model informs an online adaptive tuning engine that sets ad load under constraints
Dec 2025 · NeurIPS 2025 Workshop on Constrained Optimization for ML

Causal-Informed Hybrid Online Adaptive Optimization for Ad Load Personalization in Large-Scale Social Networks
Prompts flow through an encoder and semantic cache to a token-length predictor and a scheduler in front of the vLLM engine
Jun 2024 · Stanford University

Optimizing Semantic Caching for Sequence Prediction and Memory-Optimized LLM Inference Scheduling
An image tower paired with a BERT text tower whose weights come from transformers pre-trained on MIMIC-III and PubMed
Apr 2023 · MIDL 2023, short paper

Improving Zero-Shot Detection of Low Prevalence Chest Pathologies using Domain Pre-trained Language Models
Surface brightness profiles of two active Centaurs with their telescope images inset
May 2019 · The Astronomical Journal

Photometry of Active Centaurs: Colors of Dormant Active Centaur Nuclei

Education, work and honors

All roles

Stanford University

MS, Computer Science2023 – 2026
  • GPA 4.15
  • Vice President, CS Student Council

Harvard University

AB, Computer Science, summa cum laude2019 – 2023
  • GPA 3.99
  • Thesis with Stratos Idreos, Data Systems Lab
  • Teaching Fellow for CS161 (Operating Systems) and CS209B (Advanced Data Science)
  • President, Harvard Undergraduate Robotics

Recent roles

  • Epsilon Health, Founding Researcher2026 – now
  • Meta, Senior Machine Learning Engineer, Area Lead2023 – 2026
  • Solesca, Machine Learning Engineer (part-time)2023

Honors

  • First place, Correlation One / Citadel Boston Regional Datathon ($15,000)2022
  • Certificate of Distinction in Teaching, Harvard2022
  • Harvard Scholar Award (top 5% of class)2020, 2021
  • National AP Scholar2019