Emre Şahin

• • Served as ML Lead for Google X Team's Ship Tracking Project utilizing Distributed Acoustic Sensor (DAS) data.

• • Optimized PostGIS queries using AIS telemetry data to significantly accelerate ship positioning resolution.

• • Created signal-processing-based datasets and trained/fine-tuned custom DINOv3 models.

• • Architected production deployment solutions leveraging Cloud Run and Google Vertex AI.

• • Engineered Dataflow (Apache Beam) processing pipelines for parallel ingestion of multi-terabyte datasets.

• • Built an end-to-end cloud ingestion pipeline saving processed acoustic telemetry into Cloud SQL.

• • Developed and deployed a standalone DAS visualization and data-labeling tool using Python Panel/Holoviz on GCP.

• • Design and build end-to-end production ML and MLOps infrastructures for global enterprise clients.

• • Developed a real-time multi-camera face recognition engine using transfer learning (VGG-16, TensorFlow) synchronized via Redis IPC.

• • Created and maintained open-source MLOps state and data management tooling in Rust and Python (Xvc).

• • Developed enterprise LLM and NLP orchestration pipelines optimized for contract negotiation and legal operations.

• • Evaluated and integrated commercial and open LLMs (OpenAI, Anthropic, Gemini, Llama) for metadata extraction and clause categorization.

• • Maintained core product backends written in Ruby on Rails and high-throughput Python inference layers (FastAPI, Flask).

• • Implemented production Retrieval-Augmented Generation (RAG) using Pinecone, Qdrant, pgvector, and Sentence Transformers.