Vetted Keras Professionals

“CoreAI intern at The Home Depot who improved the Magic Apron Assistant by building a production video ingestion + RAG retrieval system for long videos (uploads and YouTube), including a graph-based retrieval module to speed up and improve relevance. Experienced with Kubernetes orchestration (HPA) and production reliability practices like caching, monitoring, regression testing, and stakeholder-driven requirements.”

“Deployed a real-time LLM-driven call center summarization and agent-assist platform at Fifth Third Bank, combining transformer models (BERT/GPT) with FastAPI inference on AKS and vector storage (ChromaDB/PostgreSQL). Emphasizes production-grade reliability (autoscaling, CI/CD, monitoring) and measurable evaluation (A/B testing), and translates model outputs into business-facing Power BI insights for call center leadership.”

“ML/AI engineer who built and productionized an NLP system at PurevisitX, orchestrating end-to-end ML workflows with Airflow (S3 ingestion through auto-retraining) and optimizing for drift and low-latency inference. Also partnered with Citibank risk teams on a fraud detection model, translating results via dashboards and iterating thresholds based on stakeholder feedback.”

“Built a production LLM-RAG system at McKesson to let internal healthcare operations teams query large volumes of unstructured operational documents via natural language with source-backed answers, designed with HIPAA/FHIR compliance in mind. Demonstrated strong production engineering across hallucination mitigation, retrieval quality tuning, and latency/scalability optimization, using LangChain/LangGraph and Airflow plus rigorous evaluation/monitoring practices.”

“Software engineer with hands-on experience building a decentralized file-sharing dApp, bridging a React frontend with Ethereum smart contracts via Web3.js and integrating IPFS for decentralized storage. Demonstrates a rigorous, measurement-driven approach to performance optimization (profiling + benchmark/regression loop) and strong ownership in high-stakes environments, including Fircosoft sanctions platform optimization and rapid production hotfixes for user-impacting issues.”

“Healthcare AI/ML engineer with Epic experience who built and deployed a HIPAA-compliant GPT-4 RAG clinical assistant over large medical document sets, emphasizing privacy controls and low-latency performance. Also automated end-to-end retraining and deployment of patient risk models using orchestration/CI-CD (Jenkins, SageMaker, MLflow), cutting deployment time from hours to minutes while improving reliability.”

“Built and deployed a production multimodal language translation platform (text-to-text, speech-to-text, text-to-speech) using fine-tuned pretrained models (NLLB, XLSR), MLflow-orchestrated pipelines, and Docker/Kubernetes on AWS. Worked closely with non-technical linguists to tackle data cleaning and dialect variation in minority languages, improving accuracy through consistent evaluation and monitoring.”

“ML/NLP engineer with healthcare and industrial IoT experience: built an Optum pipeline that converted 2M+ physician notes into structured entities and linked them with claims/pharmacy data to create an actionable patient timeline. Deep hands-on expertise in production NER, entity resolution, and hybrid search (Elasticsearch + embeddings/FAISS), plus robust data engineering practices (Airflow, Spark, data contracts, auditability) and experimentation-to-production rollout via shadow mode and feature flags.”

“ML/AI engineer who deployed a real-time, edge-based computer-vision pipeline for produce recognition in retail self-checkout to reduce shrink. Demonstrates strong end-to-end production chops: multi-camera data calibration/sync, ranking-based modeling for fine-grained classes, latency-focused optimization, and continuous A/B testing/monitoring with guardrails. Experienced with ML orchestration (Kubeflow Pipelines, Airflow) and CI/CD via GitHub Actions, and collaborates closely with store operations to make interventions usable in the checkout flow.”

“Robotics/AV perception engineer who built a semantic-segmentation road detection system and integrated it into a ROS-based real-time pipeline (ROS bag camera feed to live monitor) achieving ~12 FPS. Strong in practical deployment work: solved multi-library versioning issues (ROS/OpenCV/TensorFlow), containerized the stack with Docker, and optimized inference by shifting runtime to C++ for large latency gains on NVIDIA hardware.”

“Built an Arduino-based obstacle-avoiding robot using sonar/laser sensors and improved performance from 0.60 to 0.87 accuracy through sensor-fusion thresholding and iterative tuning. In an internship, optimized a legal-document NLP pipeline by switching to a distilled/quantized transformer and offloading inference to a GPU-backed Flask service, cutting inference time by 40%+ without added infrastructure spend.”

“Robotics/perception engineer who led and owned ARC, a cooperative perception system for autonomous vehicles that aligns and fuses multi-vehicle LiDAR point clouds in real time. Built a ROS-based multi-node pipeline with grid-based spatial reasoning and motion-compensated data sharing, achieving <20 ms compute latency and sub-7 cm alignment error; accepted to ACM SenSys 2026.”

“AI Automation Engineer at BalancedTrust who single-handedly shipped production LLM features for FinTech compliance: a policy gap-analysis pipeline (SOC 2/GDPR) and a RAG-based regulatory chatbot. Deeply focused on reliability in high-stakes legal/compliance settings, with strong production engineering (edge functions, parallelized batching to cut latency, structured JSON outputs, guardrails, and monitoring) and close collaboration with non-technical compliance experts.”

“Built and deployed an enterprise RAG application at Centene to help clinical teams retrieve insights from large internal policy document sets, cutting manual research by 30–40%. Implemented custom domain-adapted embeddings (SageMaker + BERT transfer learning) and hybrid retrieval (BM25 + Pinecone) to drive a 22% relevance lift, and ran the system in production on AWS EKS with CI/CD, MLflow, and Prometheus monitoring (99% uptime, ~40% latency reduction).”

“Healthcare data/NLP practitioner with experience at UnitedHealthcare building production ML systems that connect unstructured call center transcripts and medical notes to structured claims data. Has delivered measurable impact (25% classification accuracy lift; ~30% relevance improvement) using classical NLP, embeddings (Sentence-BERT + FAISS), and AWS SageMaker deployments with robust validation and drift monitoring.”

“Post-graduate, self-directed robotics/RL practitioner who independently built a modular reinforcement learning training framework in Python using Stable-Baselines3, Gymnasium, and PyTorch. Emphasizes reproducible experimentation (multi-seed validation), simulation (PyBullet/Box2D), and systematic comparison of algorithms/environments via a factory-pattern architecture.”

“Data/ML practitioner with experience at Humana and Hexaware, focused on turning messy, semi-structured datasets into production-ready pipelines. Built an age-prediction model from book ratings using heavy feature engineering and multiple regression models, and has hands-on entity resolution (deterministic + fuzzy matching) plus embeddings/vector DB approaches for linking and search relevance.”

“Built Hexagon’s production Text-to-CAD Copilot that converts text and rough sketches into editable CAD code, combining GraphRAG (Neo4j/LangChain) with a Gemini-powered vision module and multi-agent geometric validation—cutting manual modeling from a day to ~45 seconds and driving retrieval latency below 50ms. Also has large-scale GCP data/ML orchestration experience (Airflow/Cloud Composer, Dataflow, Pub/Sub, Snowflake) processing 50M+ daily records with drift monitoring and automated reliability controls.”

“Defense-domain AI engineer who built a production ReAct-style RAG system for military training data/material generation, scaling to ~1000 users and cutting generation time by 50%. Also has experience designing GPU-cluster parallel computation with PyTorch and handling production incidents involving database performance and schema design.”

“ML engineer with General Motors experience deploying production AI systems, including a BERT-based sentiment classifier for over a million customer support call transcripts (reported ~91% precision) and sub-200ms latency via FastAPI/Docker optimization. Also built predictive maintenance models and automated retraining/monitoring workflows using Airflow and MLflow, collaborating closely with non-technical customer support stakeholders.”

“Built and productionized an LLM-based financial document analysis system using a RAG pipeline, including robust ingestion/chunking/embedding workflows, vector DB retrieval, and an AWS-deployed FastAPI service containerized with Docker. Demonstrates strong applied expertise in improving retrieval quality and latency at scale, plus hands-on experience debugging agentic/LLM workflows with monitoring and trace-based analysis while supporting demos and customer-facing adoption.”

“Early-career cloud/appsec-focused engineer with hands-on experience building secure, observable microservice systems on AWS (IAM least privilege, KMS encryption, Secrets Manager, CloudWatch, ALB) and troubleshooting autoscaling-related 500s down to connection pooling issues. Also deployed heavy ML workloads on Kubernetes by decomposing diffusion/transformer services, using workload identity to eliminate static credentials, and maintaining GitOps-style deployment audit trails.”

“Built and productionized a RAG-based LLM research assistant for biomedical and regulatory document search using Mixtral 7B on SageMaker, LangChain, and Milvus, cutting research time by ~40%. Has hands-on multi-cloud MLOps experience across AWS/Azure/GCP with Kubeflow/Airflow/Composer plus Terraform + ArgoCD, and applies rigorous evaluation/monitoring (latency, accuracy, hallucinations). Also partnered with a non-technical PM to deliver an insurance policy Q&A chatbot that reduced customer response time by 30%+.”