Akshat Shrivastava

Experience

Facebook

Research Scientist

Working at Meta AI FAIR, my research includes pre-training, LLMs, and multimodal modeling. My work includes:

  • Efficient Adaptive Multimodal Architectures (MoMa)
  • Efficient LLM Inference (LayerSkip)
  • Multimodal Foundation Models (Chameleon)

 

Experience

Facebook

Research Scientist

Working on Language AI in AR AI, where my research has shipped in products such as Rayban Meta, Rayban Stories, Oculus, and Portal.

My work includes:

  • On-device LLMs
  • LLM Strategy for AR Devices
  • Non-autoregressive generation
  • Audio based NLP Systems for Spoken Language Understanding
  • Data efficient and high quality semantic parsing

 

Experience

Interactive Data Lab

Graduate Research Assistant

  • Working on Natural Language Processing in Exploratory Data Analysis
  • Advisor: Jeffrey Heer

 

Experience

Facebook

Software Engineer Intern

  • Working on the News Feed Content Classifcation Team
  • Worked on Text Modeling, Model Calibration, and Feature Discovery Tools
  • Work invovled Python, C++, Hack (PHP), and Caffe2

 

Experience

Microsoft

Software Engineer Intern

  • Working on the Microsoft Intune Authorization team
  • Presented and iterated on service design and security
  • Implementing a protocol for authorizing service requests in oAuth in C#

 

Experience

Facebook

Software Engineer Intern

  • Working in New York City on the Service Lab: Performance Testing for Backend Servers
  • Collaborated with multiple teams across Facebook, Presented design decisions
  • Responsibilities included developing in Python backend, and front end work in PHP and React.js

 

Experience

Interactive Data Lab

Undergraduate Research Assistant

  • Worked on Data Flow Optimization in Vega-Lite (a visualization grammar) to reduce redundancy and the size of the data flow graph
  • Worked on Implementing a Window transform in Vega-Lite to calculate aggregates for items in a defined window
  • Worked on various features for the constraint engine for Voyager (a visualization recommendation tool)
  • Vega-Lite
  • CompassQL (Voyager Constraint Engine)

 

Experience

Google

Software Engineer Intern

  • Working on the payments infrastructure team, worked on developing generators for APIs
  • Implemented experiments in buy flows to test for conversion rates
  • Primarily worked in backend Java and with Protocol Buffers

 

Experience

Tableau

Software Engineer Intern

  • Worked on Tableau Runtime to develop tools and algorithms for validation and visualization of Tableau Runtime artifacts
  • Presented and participated in design reviews and hackathons
  • Majority of work consisted of manipulating graphs and trees, primarily done in Python/DJango and HTML/CSS/JS

 

Experience

Caradigm

Software Engineer Intern

  • Developed enterprise software for health care. Built components that displayed various health data for patients
  • Collaborated and contributed with team to improve and add functionality to applications
  • Used and enhanced my skills in front-end web development, by using JavaScript, Knockout, JQuery, CSS/HTML5, and backend development using SQL