Headline

AI Solution Architect @ MPCL | Scalable AI Systems | MLOps | Kubernetes, Docker, Linux | CUDA & NVIDIA Stack

About

AI Solution Architect / AI Engineer with hands-on experience designing and deploying scalable, production-grade AI systems across GPU, Kubernetes, and cloud-native environments. I specialize in Generative AI, RAG systems, and GPU-accelerated inference, with strong expertise in Python, PyTorch, CUDA, TensorRT, Docker, and Kubernetes. At Micropoint Computers, I’ve built and optimized end-to-end AI pipelines - deploying multimodal LLMs on Kubernetes, improving GPU utilization, and reducing inference latency at scale. My work spans from model development to MLOps/AIOps, including monitoring, performance tuning, and post-deployment support for real-world AI systems handling high request volumes. I also bring a strong data background, leveraging SQL, Power BI, and analytics to support data-driven decision-making. Beyond project work, I actively contribute to open-source projects, solve SQL and ML challenges, and conduct Linux & NVIDIA stack training sessions to help teams adopt AI systems confidently. Always eager to tackle challenging problems in AI systems architecture, GenAI, MLOps, and cloud-native ML.

Location

Pune Division

AI stack

anthropic apis

claude

hugging face

kimi

notebooklm

ollama

openclaw

opencode

Work experience

Solution Architect - AI at Micropoint Computers Pvt. Ltd., Nov 2024 - Present · 1 yr 9 mos, Pune District, Maharashtra, India, • Designed and deployed GPU-accelerated AI inference systems using NVIDIA CUDA, cuDNN, TensorRT, improving inference latency by 30-45% on production workloads. • Built production-grade RAG pipelines with LLMs and vector databases, deployed on Kubernetes, supporting 1K+ daily inference requests. • Containerized AI workloads using Docker and orchestrated them on Kubernetes, implementing GPU scheduling and scaling to improve GPU utilization by ~25%. • Optimized deep learning models through TensorRT conversion and CUDA tuning, increasing throughput by 1.5–2× on NVIDIA GPUs. • Implemented MLOps/AIOps practices (monitoring, logging, performance tuning) for reliable, low-latency AI systems in production.

Data Analyst at WeAgile Software Solutions Pvt. Ltd., May 2024 - Aug 2024 · 4 mos, Pune Division, Maharashtra, India, • Optimized complex SQL queries and ETL pipelines processing 5M+ records, improving data extraction efficiency by 10% and accelerating business reporting cycles by 2 days per sprint. • Developed and maintained interactive Power BI dashboards integrating 3 data sources, enhancing business insight delivery by 12% and supporting data-driven decisions for 20+ stakeholders.

Education

Bachelor of Technology - BTech, IT - Data Science, Ajeenkya D Y Patil University

Elementary Education, kendriya vidyalaya

Public links

github: https://github.com/18Tushar