Pallavi Raiturkar

Staff Software Engineer · GitHub

Pallavi Raiturkar

I started in eye-tracking research, spent five years in full-stack engineering, moved into AI systems, and now lead technical work on Copilot Chat at GitHub. Across all of it I have built the same way: pay attention to what people actually do, then design for that.

View selected work Email me

Selected work

Systems, outcomes, and the product questions behind them. These are selected from my recent engineering roles.

GitHub

Staff Software Engineer

Jul 2025–Now

Making Copilot Chat faster, simpler, and easier to trust.

I lead technical work across Copilot Chat on GitHub.com: architecture, quality, performance, agent capabilities, and the platform beneath the experience.

40% lower P95 time to first token through telemetry-led performance work spanning flamegraphs, token reduction, attachment hydration, and parallel tool execution.

Re-architected multi-agent workflows into skill-based systems and built offline LLM-as-a-judge evaluations for faster, safer iteration.

Walmart

Staff Software Engineer

May 2024–Jul 2025

Building agents for decisions, not demos.

I developed and scaled enterprise assistants for data insights, workforce planning, feedback analysis, and recommendations, then built the backend systems needed to run them reliably.

A LangGraph text-to-SQL and insights agent served 10K+ daily queries and reduced resolution time by 35%.

A workforce planning agent cut approval time by 50% and automated more than 200 hours per quarter.

Walmart

Software Engineer II → Senior Software Engineer

Sep 2019–May 2024

Learning the systems underneath the interface.

My path at Walmart moved through data pipelines, full-stack applications, real-time dashboards, machine learning, deployment frameworks, and evaluation. Working across layers made product quality inseparable from system quality.

Built grocery fulfillment applications with Python, Angular, and React, plus real-time operational dashboards using WebSockets and Elasticsearch.

Led Kubernetes deployment and testing frameworks that improved release speed, rollback safety, and production observability.

Research made user empathy operational.

In human-centered computing, I used eye tracking, pupil response, physiology, and machine learning to study what people attend to and how they respond.

That work taught me to care about what users want in a way I still practice as an engineer: observe behavior, measure the real experience, and let the evidence change the product.

University of Florida · Research Assistant · 2015–2019

  • Eye tracking across 2D video and virtual reality
  • Visual perception of synthetic and photographic portraits
  • Physiological response and emotional engagement
  • Gesture recognition for human-computer interaction

Publications

Peer-reviewed papers, conference work, and technical reports.

  1. 2019 International Journal of Semantic Computing A Benchmark of Four Methods for Generating 360° Saliency Maps from Eye Tracking Data
  2. 2019 Society & Natural Resources Using Audience Physiology to Assess Conservation Messages and Animal Taxa
  3. 2018 ACM VRST An Evaluation of Pupillary Light Response Models for 2D Screens and Virtual Reality HMDs
  4. 2018 University of Florida Technical Report Identifying Computer-Generated Portraits: An Eye Tracking Study
  5. 2016 ACM Symposium on Applied Perception Decoupling Light Reflex from Pupillary Dilation to Measure Emotional Arousal in Videos
  6. 2016 ACM Symposium on Applied Perception Scan Path and Movie Trailers for Implicit Annotation of Videos
  7. 2016 ACM Symposium on Applied Perception Measuring Viewers’ Heart Rate Response to Environment Conservation Videos

Studio, off the clock.

I like exploring across forms and media. This is a separate track from my engineering work: a place for making, looking, and learning without needing the experiment to become a product.

The studies here are original, site-specific generative pieces that establish the visual room while the personal collection grows.

Abstract mineral-green contour study on a dark field
Form studyContour / repetition
Overlapping iris and mineral color planes
Color studyBalance / interference
Textured dark surface with hand-drawn light marks
Material studyGrain / gesture
Dark geometric form cut by a pale beam of light
Light studyShadow / aperture

Let’s work on something that matters to the people using it.

I’m interested in senior engineering roles, technical leadership, AI product work, research-informed development, and thoughtful collaborations.

Email LinkedIn GitHub