Nirmal Senthilkumar
Computer Science @ Purdue University
Specializing in GPU Programming, Machine Learning, and Computer Graphics
Education & Certifications
Purdue University
B.S. in Computer Science
Aug 2023 – May 2027
Concentrations:
- • Machine Intelligence
- • Computer Graphics and Visualization
Key Coursework:
AWS Certified Cloud Practitioner
January 2025
Foundational knowledge of AWS cloud services and cloud computing, including core AWS services, security, architecture, pricing, and support.
Professional Experience
Undergraduate Researcher
STyGIANet Research Lab
- ▸Researching MoE (Mixture of Experts) model training with focus on token routing and network optimization
- ▸Developing token bandwidth minimization techniques for Tree of Thoughts models
- ▸Working on next-generation photonic interconnects for GPU clusters to break bandwidth barriers in distributed training
GPU & ML Intern
Advanced Micro Devices (AMD)
- ▸Built out ADAS (Advanced Driver-Assistance Systems) using AMD Versal SOCs (CPU, GPU, FPGA, NPU)
- ▸Processed images using PL, ran inference on NPU (AIE) with YOLOX/YOLOv8/MobileNet (inference:<5ms)
- ▸Contributed to Vulkan/OpenGL SDKs for x86/ARM systems running Wayland/X11 for Radeon/ARM Mali GPUs
- ▸SDKs included demos of: ray tracing, n-body particles, occlusion culling, GEMM, Convolutional/Gaussian Filters
Undergraduate Researcher/Subteam Lead
Thermo Fisher x The Data Mine
- ▸Developed time-series forecasts using machine learning models, generating visual projections for a 3+ year outlook
- ▸Worked in the Machine Learning subteam in an AGILE scrum project management environment
- ▸Presented findings and methodologies used in a symposium filled with dozens of industrial experts
Robotics/Software Developer Intern
Nod.ai
- ▸Utilized ROS for controlling robots that emulate human movement inputted through a video stream
- ▸Dissected video streams using OpenCV and CMU's Open-Pose to route commands to ROS Publisher/Subscriber
- ▸Accelerated video-processing to be real-time using Nvidia CUDA on a Jetson TK1 board running Linux
Technical Skills
Languages
Full-Stack
Graphics/ML
Methodologies
Featured Projects
A showcase of my work in GPU programming, full-stack development, and ML infrastructure
Wave: ML Kernel Development
Contributing to Python DSL for high-performance machine learning with GPU kernel implementations
- ▸Contributing GPU kernels for attention mechanisms to Wave, a Python domain-specific language for ML
- ▸Implementing high-performance compute kernels that compile to optimized GPU code via MLIR
- ▸Developing kernel optimizations for transformer architectures and attention operations
GPU Rock-Skipping Simulation
Realistic rock skipping simulation with custom GLSL shaders for dynamic water ripples and caustics
- ▸Designed realistic rock skipping simulation using custom GLSL shaders for dynamic water ripples and caustics
- ▸Implemented real-time trajectory calculations + preview line, utilized Laplacian equations for wave dynamics
- ▸Applied various lighting effects - normal maps, shadow maps, ray tracing - to create breathtaking/efficient visuals
www.motion.lat
Full-stack geographical social media app with React frontend and Express backend
- ▸Developed full-stack geographical social media app using React frontend + Express backend + Google Maps API
- ▸Architected efficient hybrid MongoDB + PostgreSQL database for efficient data access alongside Prisma ORM
- ▸Deployed backend services via Docker on GCP with automatic CI/CD pipeline with Scrum project management
Llama 3.1 70B Serving on Kubernetes
Scalable Kubernetes cluster for LLM inference with OpenAI-compatible API
- ▸Implemented a scalable, replicating K8s cluster to handle inference requests exposing an OpenAI Endpoint
- ▸Deployed a baremetal cluster with kubeadm running VLLM containers hosting Llama-3.1-70B for chat completions
- ▸Utilized Istio service mesh to handle container networking through TCP/IP, gRPC, HTTP and TLS protocols
Get In Touch
I'm always open to discussing new opportunities, collaborations, or just chatting about technology. Feel free to reach out!
senthi22 [at] purdue [dot] edu
GitHub
github.com/nirmie
linkedin.com/in/nirmal-senthil
Looking for a passionate developer with expertise in GPU programming, ML, and full-stack development?
Send Me an Email