Hi, my name is
Gaurav Batra.
I build scalable AI systems.
I'm a Master's student at UW-Madison specializing in Systems and AI/ML, with 3+ years of experience building production MLOps platforms and optimizing ML infrastructure.
Currently seeking full-time opportunities in AI Infrastructure, MLOps, and Distributed Systems for 2026.
About Me
Hello! I'm Gaurav, a Master's student in Computer Science at UW-Madison (GPA: 3.92/4), specializing in systems, distributed computing, and AI/ML.
I bring over 3.5 years of professional experience as an AI Platform Engineer at Couture.ai, where I built scalable MLOps infrastructure, and as an AI intern at NVIDIA, optimizing video compression with deep learning. Most recently, I'll be joining Sigma Computing as a Software Engineering Intern in Summer 2025.
I graduated with honors (CGPA: 9.48/10) from IIIT Hyderabad, where I researched bandit algorithms at the Machine Learning Lab under Prof. Naresh Manwani, publishing my work at PRICAI'21.
Beyond my academic and professional pursuits, I'm passionate about gaming. My favorites range from strategy classics like Age Of Empires to action-adventure titles like Assassin's Creed.
Here are a few technologies I've been working with recently:
- Python, C++, Golang, Rust
- PyTorch, TensorFlow, Hugging Face
- Kubernetes, Docker, AWS, GCP
- FastAPI, PostgreSQL, Redis
- Spark/PySpark, FUSE, Bash
Where I've Worked
Software Engineering Intern - AI @ Sigma Computing
May 2025 - Aug 2025
- Built and launched a semantic search feature in Ask Sigma, allowing natural language queries to find workbooks, driving a 30% improvement over Sigma's older search system.
- Deployed the feature as a scalable service on Kubernetes, optimizing for performance and reliability, resulting in 100+ daily queries served with 10% lower latency.
- Partnered closely with the VP of AI and Co-founder to deploy the feature in production and showcase it in Sigma's 2025 AI product launch.
- Tech Involved: Python, Kubernetes, Semantic Search, NLP, Production ML Systems.
AI Platform Engineer @ Couture.ai
June 2021 - July 2024
- Core platform engineer responsible for designing, building, and scaling Couture.ai MLOps platform to manage the end-to-end machine learning lifecycle.
- Optimized legacy pipelines by writing latency-critical components in Rust, achieving 40% reduction in deployment time and 20% improvement in inference throughput.
- Spearheaded deployment of a secure, on-prem MLOps stack for the platform's first U.S. client on AWS EKS, driving enterprise compliance and contributing to $250k ARR.
- Engineered 15+ microservices in Python (FastAPI) and Golang, deployed with Kubernetes, and automated MLOps workflows with CI/CD across AWS and GCP, reducing manual deployment by 90%.
- Designed core architectures, including model registry, connector modules, and long-term roadmap for multi-tenant MLOps capabilities.
- Mentored two interns in developing a new ML observability stack using Grafana Loki, reducing critical production issue discovery time by 75%.
- Tech Involved: Python, Golang, Rust, Kubernetes, AWS, GCP, FastAPI, PostgreSQL, Redis.
Software Engineer Intern - AI @ NVIDIA
April 2020 - June 2020
- Engineered a TensorFlow-based Convolutional Neural Network for optimized image compression, inspired by Better Compression with Deep Pre-Editing.
- Achieved a 15% improvement in Peak Signal-to-Noise Ratio scores and a 10% increase in SSIM metrics, outperforming traditional JPEG methods.
- Optimized video encoding by reducing bits-per-pixel to maintain stable frame rates under bandwidth constraints.
- Conducted extensive experiments and model tuning, leading to 27% performance improvement.
- Tech Involved: TensorFlow, PyTorch, Computer Vision, Image Compression.
Undergraduate Researcher @ Machine Learning Lab
May 2019 - August 2021
- Authored the paper "Multiclass Classification under Dilute Bandit Feedback", published at PRICAI'21.
- Showed that ML models trained with partial feedback can achieve accuracy comparable to full-feedback methods.
- Research focused on classification algorithms in the Bandit and Semi-Bandit Settings under the guidance of Prof. Naresh Manwani.
- Designed algorithms for multiclass classification under bandit setting with dilute feedback, inspired by the Banditron algorithm.
- Tech Involved: Python, Machine Learning, Bandit Algorithms, Research.
Research Intern @ Virtual Labs (Vlead)
Novemeber - December 2018
- In Fall of 2018, I worked as a research intern at Virtual Labs, IIIT Hyderabad, a social initiative of the Government of India.
- I developed full fledged experiments and interactive artefacts for various data structures and algorithms using software engineering research principles.
- Me and my friend Abhinav worked togther on the project and presented our work at the RnD Showcase 2019 [Poster].
- Tech Involved: Javascript, HTML & CSS, three.js.
Trainee @ Docturnal
August - November 2018
- Docturnal is a is a point of care screening and diagnostics provider of non-invasive and proactive detection of diseases.
- It's flagship product is Timbre, which is used to screening and detection of Tuberculosis in patients.
- I was a part of the 5 member team, responsible for designing the UI of the web-app for Timbre.
- I was responsible for automating the process of retrival of patient data and delivering it to relevant personnel like doctors in clinics, hospitals.
- Tech Involved: javascript, postgresql, django.
Some Things I've Built
Featured Project
Cshell
Cshell is an interactive user defined bash-like shell that supports semi-colon separated list of commands. It supports some custom commands like "remindme","clock" etc. Signals like Ctrl+Z, Ctrl+C have been handled.
- C
- Operating Systems
- Threads
- Signal Handling
Featured Project
Reinforcement Learning Algorithms
In this project, I have used Monte-Carlo Methods and Temporal Difference Learning on couple of games and toy problems.
- Trained an agent that plays the Tic-Tac-Toe using Monte-Carlo Methods.
- Trained an agent that generates the optimal policy through TD-Methods in the Frozen-Lake Environment.
- Built a DQN which can play the cart-pole game.
- Monte-Carlo Methods
- Temporal Difference Algorithms
- Deep Q-Learning Network
Featured Project
Airplane Simulator
It is a 3D emulation of a jet fighter plane game built in OpenGL. This game was made as a part of the Graphics Course. This game has been developed keeping the physics involved in airplane movement in mind and supports multiple camera angles.
- OpenGL
- Game Design
- Cpp
Other Noteworthy Projects
view the archiveDeduplication Pipeline & Multi-GPU LLM Fine-Tuning
Developed a high-throughput deduplication pipeline using conventional hashing and MinHash to eliminate exact and near-duplicates from over 10M documents across GitHub and Common Crawl sources. Built a multi-GPU fine-tuning pipeline with Hugging Face Accelerate and Transformers to efficiently fine-tune LLMs featuring optimized I/O, cluster-level parallelism, checkpointing, and reproducible W&B logging.
Custom Filesystem using FUSE
Designed and implemented WFS, a custom block-based filesystem using the FUSE framework, supporting basic filesystem operations and RAID 0/1 configurations to enhance fault tolerance and performance. Built entirely in C with low-level block management.
What's Next?
Get In Touch
I'm currently seeking full-time opportunities in AI Infrastructure, MLOps, and Distributed Systems starting in 2026. Whether you have an opportunity, want to collaborate on a project, or just want to chat about ML systems and distributed computing, feel free to reach out!
My inbox is always open, and I'll do my best to get back to you!