Ali Akram

Architected enterprise Generative AI and agentic AI platforms using Azure OpenAI and other LLM providers, leveraging LangChain/LangGraph, LlamaIndex, and multi-agent architectures for knowledge management and workflow automation. Designed MCP-based integrations connecting LLMs and agents to enterprise APIs and knowledge repositories. Built production-grade RAG systems with hybrid/semantic retrieval, embeddings, reranking, and knowledge-graph approaches (e.g., GraphRAG with Neo4j). Implemented secure LLM APIs and microservices (FastAPI, Docker, Kubernetes/AKS, Azure Functions) with guardrails such as prompt-injection detection, PII masking, RBAC, audit logging, and secure tool execution. Delivered enterprise LLMOps/MLOps pipelines (prompt/model versioning, evaluation, deployment, rollback, monitoring) using MLflow and cloud-native CI/CD, along with observability via Langfuse/OpenTelemetry/Prometheus/Grafana/Azure Monitor. Partnered with product, architecture, security, and domain stakeh

Architected production ML platforms for credit risk, underwriting analytics, fraud detection, risk profiling, lending decision automation, and forecasting. Built predictive models (XGBoost, LightGBM, Random Forest, Gradient Boosting, Logistic Regression; TensorFlow/PyTorch) and implemented recommendation systems using clustering and behavioral analytics. Designed scalable inference on AWS (SageMaker endpoints, Lambda, ECS/EKS, API Gateway, FastAPI) for real-time and batch workloads. Built data engineering pipelines using AWS Glue, PySpark, S3, and Redshift. Developed production-grade RAG pipelines with Bedrock, OpenSearch/vector search, and grounded responses. Implemented LLM routing/model selection for latency and cost optimization, PEFT/LoRA fine-tuning workflows, and evaluation frameworks (RAGAS, DeepEval, Langfuse, LLM-as-a-judge) with metrics for faithfulness, relevance, retrieval quality, hallucinations, and latency/cost. Strengthened compliance needs with explainability (e.g., S

Designed scalable ML solutions on GCP for forecasting, fraud detection, budget optimization, and financial planning using Python, scikit-learn, and TensorFlow (via Google AI Platform/Vertex AI). Built ML pipelines using BigQuery, Cloud Storage, Dataproc, Cloud SQL, MLflow, and Docker. Developed classification/regression/anomaly detection/forecasting models including Random Forest, XGBoost, and Decision Trees, with large-scale feature engineering using Spark/PySpark. Implemented reusable connectors and REST API integrations for Power Platform and external enterprise/SaaS systems. Built enterprise generative AI and AI/ML platforms on GCP leveraging Gemini, BigQuery, Cloud Storage, Dataflow/Dataproc, Cloud Run, and Cloud Functions. Strengthened security and governance using GCP IAM, VPC Service Controls, Secret Manager, Cloud KMS, audit logging, and least-privilege controls. Deployed models to GKE with automated CI/CD (Cloud Build, MLflow, Docker) and implemented streaming/event-driven so