Maharsh Hetal Gheewala

Software Engineer

San Jose, California, United States7 yrs 6 mos experience
Most Likely To SwitchHighly Stable

Key Highlights

  • Reduced speaker mis-attribution from 30% to 1%
  • Engineered knowledge graph infrastructure for real-time personalization
  • Prototyped applied AI projects and contributed to open-source ML tooling
Stackforce AI infers this person is a Robotics Engineer specializing in Applied Machine Learning and Cloud Computing.

Contact

Skills

Core Skills

Applied Machine LearningKnowledge GraphsCloud ComputingData PrivacyFront-end Development

Other Skills

Speaker IDRetrieval-Augmented Generation (RAG)Embedded MLComputer VisionAmazon Web Services (AWS)Android DevelopmentSDK developmentReact.jsReact NativeDeep LearningData ScienceSoftware DevelopmentAlgorithmsCascading Style Sheets (CSS)JavaScript

About

I’m a Software Engineer II at Amazon Lab126, where I build large-scale, event-driven backend and ML systems for Amazon Astro, the home robotics platform. My work focuses on real-time identity resolution, speaker attribution, and knowledge graph infrastructure that allow Astro to understand who’s interacting, where they are, and what context it’s in—all under strict latency and reliability budgets. I’ve designed and deployed pipelines that process 300+ TPS across distributed household devices, integrating multimodal classifiers (voice, vision, session) to reduce speaker mis-attribution from 30% to 1%, and cut cross-user memory corruption by 95%. I also delivered the Artifact Synchronization Infrastructure, powering Floor Plan, Wi-Fi map, and Hangout Spot generation through AWS services like S3, SNS, and SQS. Working closely with the Visual Perception Science team, I fused high-frequency vision signals (every 200 ms) with spatial and presence data to support downstream Mobility and Sentry systems, improving entity consistency and reducing false positives by 90%. I’ve also architected a hybrid in-memory and on-disk knowledge graph layer optimized for eMMC wear (~100 MB/day) while maintaining real-time query performance. I enjoy the intersection of backend engineering, applied ML, and robotics—turning perception outputs into intelligent, stateful systems that behave predictably in the real world. Beyond work, I prototype applied AI projects, contribute to open-source ML tooling, and study how responsible AI principles can be embedded directly into product architecture.

Experience

7 yrs 6 mos
Total Experience
3 yrs 9 mos
Average Tenure
4 yrs 8 mos
Current Experience

Amazon lab126

2 roles

Software Dev Engineer II

Promoted

Dec 2024Present · 1 yr 5 mos · Sunnyvale, California, United States

  • Building multi-modal ML systems and knowledge infrastructure for Amazon Astro, enabling real-time personalization and robotics intelligence on embedded devices.
  • Multi-Modal Speaker Resolution: Designed and deployed a pipeline combining voice + vision (face tracking, FoV) with calibrated score fusion → reduced false acceptance/rejection by 22% across 100K+ households.
  • Confidence Scoring Framework: Built a real-time scoring system leveraging human-presence, SSL embeddings, and prior context → drove >90% accuracy in identity attribution and enabled dynamic conversational responses.
  • RAG Personalization (Prototype): Implemented retrieval-augmented generation pipelines → cut fallback response rates by 25% in internal trials, powering context-aware personalization.
  • Knowledge Graph Infrastructure: Engineered KG storage handling 300+ TPS with <3% corruption → streamlined personalization queries for Astro.
  • Contextual Space Representation: Designed “unnamed space” handling via metadata-driven contextual labels → unlocked personalization in unlabeled environments.
  • System Resilience: Defined reboot/crash recovery logic for in-memory components → ensured instant restoration with zero user-visible downtime.
  • Data Lifecycle Management: Developed automated KG garbage collection → reduced storage bloat and improved retrieval consistency.
Speaker IDApplied Machine LearningKnowledge GraphsRetrieval-Augmented Generation (RAG)Embedded MLComputer Vision

Software Engineer I

Sep 2021Dec 2024 · 3 yrs 3 mos · Sunnyvale, California, United States

  • Developed perception and platform features for embedded ML systems, enhancing robustness and user experience.
  • Improved ML-based pet detection by fusing computer vision outputs with presence APIs, reducing false positives by ~90%.
  • Collaborated with privacy and security teams to implement data sync and retention policies, ensuring compliance with stringent requirements.
  • Contributed to Amazon Lab126's mission of innovating smart devices, focusing on privacy and user-centric design.
Amazon Web Services (AWS)Cloud ComputingAndroid DevelopmentSDK developmentData Privacy

Telnyx

Software Engineer

Jul 2021Aug 2021 · 1 mo · Chicago, Illinois, United States

  • Real-Time Communication APIs: Built scalable backend services for high-concurrency voice and messaging APIs; optimized latency with asynchronous processing and query tuning, achieving a 20% performance gain under load.
  • Testing & Reliability: Implemented unit/integration tests for API endpoints, helping raise backend service test coverage from ~60% → ~80% during the internship period.

Superworld

Front-end Developer Intern

Jun 2020Aug 2020 · 2 mos · Chicago, Illinois, United States

  • Built and deployed a video conferencing application in React.js, improving internal communication efficiency and supporting distributed collaboration.
  • Partnered with backend engineers to host and scale web services on AWS, ensuring reliable performance during product demos.
  • Worked closely with UI/UX designers and data scientists to improve front-end performance and SEO visibility, enhancing user engagement.
  • Participated in daily Agile scrums, contributing to component integration discussions and code clean-up efforts that streamlined debugging and onboarding.
React.jsReact NativeFront-end Development

Illinois institute of technology

Community Desk Assistant

Dec 2019May 2021 · 1 yr 5 mos · Chicago, Illinois, United States

Picxele™ - perform tasks and earn

Full-stack Developer

Nov 2018May 2019 · 6 mos · Nodia, India

  • Full-Stack Development: Built and maintained features in the account management system using HTML5, Vue.js, Node.js, and SQL, improving usability and reliability.
  • Payment Portal: Designed and implemented a payment portal (integrated with Stripe APIs) to streamline customer billing and transactions.
  • Collaboration Tools: Integrated GitLab and Bitbucket version control pipelines, enabling efficient cross-team code sharing and CI/CD practices.
  • Mentorship & Onboarding: Assisted and mentored interns by debugging code, assigning tasks, and brainstorming design solutions → accelerated onboarding and reduced bug resolution time.

It association

3 roles

Board Member

May 2018May 2019 · 1 yr

Technical Team Lead

May 2017May 2018 · 1 yr

Member Of Technical Staff

Jul 2016May 2017 · 10 mos

  • Web Developer

Hj healthcare

Web Development Intern

Jun 2017Jul 2017 · 1 mo

  • Engineered websites using web development frameworks such as ReactJs, NodeJs, and MongoDB. Managed to built and successfully learned to make and deploy the website on a server

Education

Illinois Institute of Technology

Masters in Computer Science — Computer Science

Jan 2019Jan 2021

SRM IST Chennai

Bachelor of Technology (B.Tech.) — Information Technology

Jan 2015Jan 2019

Tel Aviv University

Summer Program — Cybersecurity

Jan 2018Jan 2018

Shardayatan

Higher Secondary Education — Science Stream

Jan 2014Jan 2015

Ryan International

Secondary Education — Regular/General High School/Secondary Diploma Program

Stackforce found 100+ more professionals with Applied Machine Learning & Knowledge Graphs

Explore similar profiles based on matching skills and experience