Looking for a job

Tony Malik

@tonymalik

Lead AI/ML Engineer at ANM Connections building low-latency LLM agents and RAG systems that improved retrieval relevance by 22%.

United States

Message

At ANM Connections, I lead a team building domain-specific AI agents and low-latency, cost-optimized LLM workflows. I built end-to-end RAG systems with OpenAI, Claude, and Vertex AI, improving retrieval relevance by 22% after migrating vector search to Pinecone and Elasticsearch.

Previously, at EN Smile, I designed LLM and RAG applications that reached 90% answer accuracy and cut manual support workload by 35%. At OptimSync, I trained computer vision models that achieved 93% detection accuracy in real-time use cases.

Experience

Work history, roles, and key accomplishments

Current

Lead AI/ML Engineer

Current

ANM Connections

Jan 2023 - Present (3 years 9 months)

Led a team of 5 engineers in building domain-specific AI agents and agentic workflows, focusing on low latency and cost-optimized LLM systems. Built end-to-end RAG systems using OpenAI, Claude, and Vertex AI, orchestrated with LangChain and monitored via LangSmith.

Python PyTorch TensorFlow scikit learn NLP Computer Vision LLM RAG Lang Chain Langsmith fastAPI WebSockets Pinecone Elasticsearch Crew AI Google Cloud Run GitHub Actions YOLO CLIP Vision Transformer Scrapy Selenium MLOps

Senior AI/ML Engineer

EN Smile

Jun 2020 - Dec 2022 (2 years 6 months)

Owned the design and delivery of LLM and RAG applications using OpenAI, Claude, and LangChain, achieving 90% answer accuracy and cutting manual support workload by 35%. Fine-tuned transformer models with Hugging Face for classification, entity extraction, and semantic search, improving F1 score by 16% over baseline.

Python PyTorch TensorFlow scikit learn NLP LLM RAG Lang Chain Hugging Face fastAPI Docker AWS MLFlow GitHub Actions GPU Optimization Quantization Batch Processing MLOps

AI/ML Engineer

OptimSync

Aug 2018 - May 2020 (1 year 9 months)

Trained and fine-tuned computer vision models (YOLOv5, VGG16, InceptionV2), achieving 93% detection accuracy in real-time use cases. Built end-to-end ML pipelines for data ingestion, training, validation, and deployment, reducing experiment turnaround time by 50%.