Ashraf Khan

About Ashraf Khan

Machine Learning Engineer with around 9 years of experience in building production grade scalable ML and Analytics solutions.2. Proficient in working with platforms like Azure, AWS, Databricks & Dataiku.3. Experienced in designing & building end to end machine learning products, Deploying GenAI applications, ML frameworks, Data Engineering Pipelines, training models, and architecting ML systems.4. Proficient in designing Solution Architecture, leading MLE & DS teams, collaborating with cross- functional stakeholders, and engaging with clients. Cloud Platforms:AzureAWSDatabricksDataikuNVIDIA DGX CloudMachine Learning & AI: LLMOps: RAG & Ingestion frameworks, Performance testing, LLM Evaluation, Observability, GuardrailsMLOps: Training, Inference & Drift Monitoring Pipelines, Model Deployment, Registry, Bias, explainability and governance.Machine Learning: SparkML, Sklearn, TensorFlow, TorchData Engineering: Snowflake, MongoDB, PostgreSQL, CosmosDB, Delta Lake, ADLS, S3, ETL pipelines, batch & streaming data processingDevOps & Orchestration: Docker, Kubernetes, Terraform, Azure Bicep, CI/CD (GitHub Actions, Azure Pipelines), Databricks WorkflowsObservability & Monitoring: Prometheus, Tempo, Loki, Grafana, OpenTelemetry, Azure Monitor, CloudWatch, Elastic Search

Experience

1. Lead Machine Learning Engineer

   Tiger Analytics

   Feb 2025 — Present

   Enterprise LLMOps & RAG Platform:Worked on a modular LLMOps platform on NVIDIA DGX Cloud using NIM for Chat and Embedding models, deployed on Kubernetes clusters. Designed plug-and-play RAG modules integrating LLMs (Anthropic, GitHub, OpenAI), vector DBs (Azure AI Search, Milvus), and storage systems (Azure Blob, Local) to avoid vendor lock in. Implemented observability using OpenTelemetry, Prometheus, Loki, and Tempo with Grafana dashboards. Automated LLM evaluation via LLMPerf and RAGAS. Explored NVIDIA NeMo Guardrails and adopted Guardrails AI Project for multi-stage (input, search, summarization, output) safety enforcement.2. Enterprise Data Ingestion Framework Led architecture and development of declarative data ingestion framework in Scala, successfully transitioning from Python to master functional programming and distributed systems. Engineered plugin-based solution supporting 5+ data platforms (Neo4j Graph, CosmosDB NoSQL, Delta Lake, PostgreSQL, Azure Al Search) through unified connector interface & extensible to new data platforms. Innovated Domain-Specific Languages (DSL) enabling non-technical users to define complex transformations via CSV configuration, eliminating 70% of development cycles. Built STTM processing engine with Spark Catalyst optimization supporting myriad Spark SQL transformations through expression-tree evaluation reducing custom UDF development. Implemented dual validation framework (SQL parser + registry-based) with pre-execution syntax checking, parameter validation, and comprehensive error handling preventing runtime failures. Delivered production orchestrator managing medallion architecture (Bronze-Silver-Gold) with automated lineage tracking, runtime filtering, and error recovery.

Education

• M.H. Saboo Siddik College Of Engineering

  B.E, Computer Science

  2012 — 2016