Shravani Vattem
Aiml Developer Mlops Engineer @Tata Consultancy Services
Signup · Get unlimited contacts
WORK HISTORY
Aiml Developer Mlops Engineer @Tata Consultancy Services
Santa Clara County, CA, US
Deployed and optimized Qwen2.5-VL and LLaMA models on Intel Gaudi 3 accelerators using a customized vLLM fork.• Tuned tensor parallelism, concurrency levels, and token latency to improve inference throughput and hardware utilization.• Created custom Docker entry points to handle multi-model payload execution with low memory overhead.• Developed Python tools for logging, visualizing, and comparing inference performance metrics.• Monitored disk I/O, cleaned up cache directories, and automated cron-based housekeeping for long-running jobs.• \"Optimized GPU and Gaudi workloads by profiling kernel execution, memory throughput, and latency using Nsight Systems and Habana profiling tools.\"• \"Automated deployment and scaling of inference services in Kubernetes across AWS and GCP, leveraging Helm charts and Terraform.\"• \"Implemented SLO/SLI metrics and error budget tracking in Prometheus and Grafana for real-time inference services.\"• \"Integrated Ray Serve for distributed model inference, improving throughput by X% in multi-node environments.\"• Gained hands-on experience with Intel’s Gaudi 3 architecture and Habana SynapseAI SDK for AI inference acceleration.• Aligned Unstructured AI use cases with Intel’s roadmap for edge-to-cloud LLM deployment strategies.• Worked with internal Intel software teams to resolve compatibility issues between vLLM runtime and Synapse SDK.• Validated LLM compatibility with Intel’s custom kernels, flagging unsupported ops and requesting patches.• Participated in internal verification cycles, providing structured model behavior feedback for AI stack refinement.• Collaborated with Intel engineers, Unstructured developers, and client stakeholders to align model goals and performance needs.• Participated in live triaging of model failures, environment mismatches, and performance drops during test runs.• Supported delivery timelines by preparing environments, fixing urgent issues, and contributing to demo readiness.
EDUCATION
Campbellsville University
Master of Science - MS, Computer Science
Osmania University, Hyderabad
Bachelor of Commerce - BCom
Narayana Junior College - India
Intermediate
Osmania University
Executive MBA
ABOUT SHRAVANI VATTEM
Experienced AI & ML Engineer with over 11+ years in creating and implementing intelligent, cloud-native solutions that enhance automation, customer engagement, and operational efficiency. Expertise in developing scalable chatbot applications using AWS services such as Lex, Lambda, SQS, SNS, and Event Bridge, focusing on serverless architecture and event-driven designs. Skilled in building LLM-based conversational agents and fine-tuning models like GPT-3, LLaMA, Mistral, and Falcon with methods such as LoRA, RLHF, and RAG for tailored AI experiences. Proficient in all stages of machine learning engineering, from model training and evaluation to deployment and monitoring using frameworks like TensorFlow, PyTorch, Hugging Face, and ONNX. Successful in managing real-time AI pipelines with Step Functions and Vertex AI, deploying APIs with Fast API, and integrating ML results into business intelligence tools like Power BI and Oracle Analytics. Strong knowledge of CI/CD practices, infrastructure as code, and DevOps tools including Jenkins, GitLab, MLflow, Docker, Kubernetes, and Terraform for efficient and automated AI workflows. Experienced in secure and compliant deployments using IAM policies and encryption, as well as responsible AI frameworks. Excellent at collaborating with cross-functional Agile teams to align complex business requirements with impactful AI solutions. With a background in enterprise finance and automotive AI, such as in-vehicle NLP assistants and visual defect detection systems, I offer a unique blend of domain expertise, technical skills, and enthusiasm for advancing practical AI applications. Notable Achievements: Optimized predictive model accuracy by 25%, accelerated data processing by 35% Current Responsibilities: AI model development, data analysis, cross-functional collaboration, deployment of scalable solutions Passions: AI innovation, data-driven solutions, emerging tech, and continuous learning Career Goals: To lead AI-driven projects that advance machine learning technologies and solve complex challenges
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.