I'm Anurag Maravi
Software and ML Engineer
I build end-to-end intelligent systems that connect machine learning, distributed infrastructure, and modern product engineering. My experience spans fast real-time inference, LLM and multimodal pipelines, secure cloud deployments, and full-stack development.
About Me
Summary
Passionate and results-oriented software engineer with a strong foundation in building scalable systems and intuitive tools. Skilled in full-stack development, distributed systems, and data-driven applications. Excited to contribute to challenging projects that drive innovation and enhance user experiences.
Work Experience
Software Development Engineer | Jan 2025 - Present
University of Wisconsin-Madison, TRAIL Lab
- Leading development of AI-powered education tools used across multiple research studies, including SLAI, AIBAT, RAID, Casting Lab, and WikiValue.
- Designed and deployed real-time LLM inference pipelines using Groq for fast transcription, summarization, and behavioral analysis which reduced processing time from 20 minutes to less than 3 minutes.
- Built production-ready FastAPI services and Next.js frontends for classroom-facing systems with secure GCP hosting, Firestore data modeling, and Cloud Run CI/CD.
- Developed provenance and influence scoring pipelines for Wikipedia data, aligning multi-million-token revision histories with LLM outputs for research use.
- Engineered a multiplayer collaboration module for Sherlock investigations using Firestore real-time listeners, conflict resolution logic, and synchronized chat flows.
- Implemented teacher dashboards, behavior tagging UIs, and AI-grounding systems to support transparency and responsible AI adoption in classrooms.
- Coordinated with UW DoIT Cloud Security to implement IAM hardening, privacy controls, and FERPA-aligned data storage policies.
Software Engineer Intern | May 2024 - Dec 2024
Stealth Startup
- Built advanced Figma plugins that automated design variable extraction and detached instance detection which improved designer productivity by more than 30 percent (React, TypeScript, Webpack).
- Developed a cross-platform pet feeder application with real-time device monitoring, automated scheduling, and remote food dispensing that enhanced user reliability and feature control (Flutter, WebSocket, FastAPI).
- Implemented OpenAPI-first backend development workflows that reduced client integration errors and improved cross-team iteration speed.
Research Software Engineer | Jul 2018 - Dec 2022
Carnegie Mellon University
- Developed Mites, a large-scale IoT sensing platform deployed across 314 devices in a 90,000 sq ft building enabling multi-modal data collection, real-time anomaly detection, and ML-driven activity inference.
- Built and maintained distributed backend services with Python, gRPC, and RabbitMQ which improved system reliability and reduced network overhead through opportunistic transmission strategies.
- Designed interactive dashboards for occupancy, appliance usage, and noise analytics which lowered facility response time by 30 percent and provided actionable insights to campus operations.
- Optimized sensor firmware, data pipelines, and feature extraction layers reducing CPU load by 55 percent and enabling long-term stable deployments.
- Collaborated on ACM IMWUT 2023 publication documenting system design, deployment, and evaluation methodology.
Software Engineering Contractor | Jan 2017 - Dec 2017
BlueBanyan Technologies Pvt. Ltd
- Built Node.js backend and Android apps for a state government hoarding management system improving operational efficiency for field officers by 30 percent.
- Developed a gym and fitness center booking platform with real-time availability, search, and membership flows improving user onboarding experience.
- Delivered a fleet tracking system for Lafarge with live location monitoring, route analytics, and data visualization that improved logistics team coordination by 25 percent.
Education
University of Southern California, Los Angeles | Dec 2024
Masters in Computer Science
Key Courses: Analysis of Algorithms, Web Technologies, Machine Learning for Data Science, Database Systems, Applied Natural Language Processing
Carnegie Mellon University, Pittsburgh | Dec 2018
Exchange Vistor Student
Pursued advanced research as part of an academic exchange program
SRM Institute of Science and Technology, Chennai | Dec 2018
Bachelor of Technology in Computer Science
Key Courses: Operating Systems, Computer Networks, Database Management Systems, Software Engineering
My Skills
Machine Learning Engineering
Building, training, and deploying scalable ML systems with real-time inference, monitoring, and experimentation.
LLM & Generative AI
Designing, evaluating, and deploying LLM-powered systems for education, reasoning, multimodal generation, and real-time applications.
Data Engineering & Infrastructure
Designing pipelines, managing data workflows, and optimizing infrastructure for ML and backend systems.
Full-Stack Web Development
Developing robust and responsive web applications with modern JavaScript frameworks and API-first design.
Backend & API Development
Engineering scalable APIs and backend services for real-time and data-intensive applications.
DevOps & Cloud
Deploying, scaling, and monitoring ML and web applications in cloud-native environments.
Mobile & Embedded Development
Building performant Android apps and experimenting with cross-platform development.
UI/UX Design & Prototyping
Designing intuitive user interfaces and flows with modern prototyping tools.
Testing & Automation
Ensuring software quality through unit, integration, and end-to-end testing frameworks.
Programming Languages
Fluent in writing clean and efficient code across multiple environments.
Projects
11 projects
Casting Lab – AI-Driven Scenario Authoring Tool for Teachers
November 18, 2025
Casting Lab enables teachers to create interactive, AI-powered learning scenarios with image generation, branching dialogues, and custom characters for K-12 classrooms.
SLAI – Science & Language AI Tutor for K-12 Classrooms
August 12, 2025
SLAI is an AI-powered system that helps teachers evaluate student science explanations, provide structured feedback, and scaffold language development through transparent and responsible LLM workflows.
RAID – Responsible AI Deliberation Tool for Classrooms
May 2, 2025
RAID is an AI-supported discussion platform that helps students deliberate on responsible AI dilemmas, analyze tradeoffs, and reflect on ethical considerations using structured prompts and grounded feedback.
AIBAT – AI Behavior Analysis Tool for Classroom Feedback
March 15, 2025
Built an AI-powered system that analyzes teacher–student classroom interactions, providing structured, actionable behavioral feedback for educator training and research.
Koitur – Collaborative Family Tree Platform
December 23, 2024
A work-in-progress full-stack platform to connect, preserve, and uplift the tribal community through family trees, cultural knowledge, events, and networking.
Group Streaks – Team-Based Daily Task Tracker
October 15, 2024
A collaborative streak-tracking app where teams complete daily tasks together to maintain group accountability. Built with Go, GraphQL, MongoDB, and React.
StockSearch - Real-Time Stock Analysis Platform
February 14, 2024
Created a full-stack real-time stock analysis platform with data visualization, stock trend insights, buy/sell simulation, and financial recommendations, improving user decision-making.
QuantumGPT - Data-Efficient LLM Augmentation Platform
January 10, 2024
Developed a platform to enhance GPT-4 for scientific applications, mitigating hallucinations and logic errors, and improving large language model accuracy in complex quantum physics tasks.
IdiotBox
August 19, 2023
MLIoT – Adaptive Machine Learning System for IoT
August 19, 2023
End-to-end ML system tailored for IoT environments, supporting dynamic training, serving, and retraining with policy-based optimization and real-time adaptation.
Mites - Scalable IoT System for Smart Buildings
March 28, 2023
Developed a scalable IoT sensing system for smart buildings, achieving a 94% packet delivery rate across 314 deployed devices, enhancing real-time applications in building management and occupant wellness.
Publications
3 papers
Context Matters: Data-Efficient Augmentation of Large Language Models for Scientific Applications
arXiv preprint
Dec 2023
This paper investigates challenges in enhancing the accuracy and reliability of large language models (LLMs) for scientific applications. It highlights issues such as hallucinations and logical errors in LLM outputs and explores methods to improve their performance. The paper demonstrates the importance of context relevance and introduces techniques to self-examine and improve LLM outputs.
View PaperMites: Design and Deployment of a General-Purpose Sensing Infrastructure for Buildings
Proceedings of the ACM on Interactive, Mobile, Wearable, and Ubiquitous Technologies (IMWUT), ACM
Mar 2023
This paper presents the Mites system, a scalable hardware-software loT platform for deploying high-fidelity, general-purpose sensing in buildings. It addresses the challenges of scalability, privacy, and security, demonstrating the extensibility of the system through real-world deployment in a university building and proof-of-concept loT applications.
View PaperMLIoT: An End-to-End Machine Learning System for the Internet-of-Things
Proceedings of the International Conference on Internet-of-Things Design and Implementation (IoTDI '21), ACM
Apr 2021
This paper introduces MLoT, a machine learning system optimized for diverse loT applications. It enables adaptive training, optimization, and serving of models tailored to the unique requirements of loT environments. The system supports the entire lifecycle of loT applications, ensuring scalability, accuracy, and performance while addressing environmental and resource constraints.
View PaperAwards
2 awards
Distinguished Paper Award (IMWUT Vol. 7)
The Association for Computing Machinery (ACM)
Oct 2024
The paper "Mites: Design and Deployment of a General-Purpose Sensing Infrastructure for Buildings" was selected as one of the winners of the IMWUT Vol. 7 Distinguished Paper Awards. This recognition highlights the paper's exemplary contribution to ubiquitous computing research, particularly for its innovative design, deployment, and evaluation of a scalable loT sensing system for smart buildings. The award was selected by a committee of 16 distinguished members from the IMWUT Editorial Board and community, from a pool of 205 papers published in IMWUT Vol. 7.
The Most Complete App Award (Hasura)
Freshdesk
May 2017
Won the "Most Complete App" award at the 48-hour hackathon "Save the Hacker," conducted by Freshdesk in Chennai, sponsored by Hasura. Developed an Android application named "Kisaan," which provides a platform for farmers to sell their products directly to customers. This achievement was recognized for its innovative use of Hasura services and impactful solution for bridging the gap between farmers and customers.
Biography
1995
Born in Bhilai, Chhattisgarh—a serene steel plant city that fostered a love for problem-solving among a diverse community of engineers.
2013
Moved to Kota, Rajasthan, to prepare for engineering entrance exams and discovered my passion for computer science during late-night coding sessions.
2014
Started my journey at SRM University, Chennai, where I developed Android apps, led a tech club, and built a foundation in computer science and software development.
2018
Selected for an exchange program at Carnegie Mellon University, joining an elite cohort of two students among 800+ applicants, diving into cutting-edge IoT research.
2019
Continued my research at CMU post-graduation, transforming a 90,000 sq. ft. building into a smart hub with real-time sensors and groundbreaking IoT systems.
2023
Moved to Los Angeles for a Master’s degree at USC, embracing the sunny weather while preparing to bring my skills to the forefront of the tech industry.
2025
Aspire to join a team of innovators developing cutting-edge technology and contributing to meaningful, transformative projects as a Software Engineer.
Let's Connect
Feel free to contact me if you have any questions or just want to say hi! I'm always open to discussing new opportunities and interesting projects.
