AI/Python Engineer for ML Deployment and NLP, 5+ yrs

Satish R AI/Python Developer

+1-972-***-**** *******.*****@*****.*** Linkedin Open to Relocate PROFESSIONAL SUMMARY

AI/python Engineer with 5+ years of experience building and deploying machine learning and deep learning models for financial and enterprise data. Experienced in Python-based pipelines, time-series forecasting, NLP, and neural network models, with hands-on delivery on AWS and GCP. Strong focus on scalable data processing, model deployment, and measurable business outcomes. SKILLS

Programming Languages & IDEs: Python, R, SQL, Jupyter Notebook, Antigravity Machine Learning & Deep Learning: Decision Trees, Random Forests, Naive Bayes, K-Means, DBSCAN, Principal Component Analysis (PCA), Recurrent Neural Networks (RNNs), Artificial Neural Networks (ANNs) Deep Learning Frameworks: TensorFlow, Keras, PyTorch AI & Generative AI Technologies : Lang Chain, Mistral, Google Adk, Lang Graph, Llama Index, Retrieval-Augmented Generation (RAG), LLaMA 2, GPT-4, Fast API, Fast MCP, Arize AI

Natural Language Processing (NLP): BERT, RoBERTa, Named Entity Recognition (NER), NLTK, SpaCy, Claude Cloud Platforms: AWS (SageMaker, Lambda, CloudWatch, CloudFormation), GCP (Vertex AI, Big Query, Google Cloud AI) Libraries & Tools: Scikit-learn, Pandas, NumPy, SciPy, PySpark, Matplotlib, Seaborn, OpenCV Data Visualization & Databases: Tableau, Power BI, SQL Server, PostgreSQL, MongoDB, Vector Databases (Pinecone, FAISS, Milvus) MLOps & Deployment: Docker, Kubernetes, CI/CD Pipelines, RESTful APIs WORK EXPERIENCE

Apple(contract) USA

AI/python developer Aug 2024 – Present

Built Python-based ANN pipelines to process and index petabyte-scale audio embeddings for large-scale music platforms, enabling fast similarity search, recommendation, and content discovery across billions of tracks and reducing manual curation and validation effort by 40%.

Implemented RoBERTa-based Transformer NER models to extract artist and label entities, genres, moods, explicit-content indicators, and licensing metadata from audio transcripts and catalog text, achieving 94% extraction accuracy and significantly improving catalog enrichment and content compliance workflows.

Developed GenAI-driven Retrieval-Augmented Generation (RAG) solutions using LlamaIndex with Pinecone and FAISS to enable semantic search and contextual insights across music catalogs, content policies, and licensing guidelines, reducing analyst and curator research time by 35%.

Integrated large language models (LLaMA 2, GPT-4, Gemini, and Mistral) via LangGraph and Google ADK to orchestrate multi-step agentic knowledge workflows, enforcing role-based access control, auditability, and model governance aligned with enterprise content and compliance standards.

Built and exposed scalable AI services using FastAPI and custom MCP (Model Context Protocol) servers, enabling secure access to embedding search, RAG retrieval, and agent tool execution, supporting seamless integration with ingestion, recommendation, and content moderation systems.

Designed and implemented an LLM and agent evaluation (Evals) framework with multi-turn regression testing, tool-use trajectory scoring, golden datasets, and CI/CD quality gates, preventing regressions in retrieval quality, tool selection accuracy, and overall task success rates.

Implemented agent guardrails and security controls including tool schema validation, allowlisted actions, least-privilege credentials, prompt-injection defenses, PII/content filtering, and human-in-the-loop approvals, ensuring safe, policy-compliant agent execution in production environments.

Built a stateful agent runtime with memory and persistence mechanisms (checkpoint/replay, session state, retrieval memory), along with timeouts, retries, streaming responses, and fallback logic, improving reliability for long-running music workflows such as catalog updates and moderation queues.

Trained TensorFlow-based ANN models for listener behavior and playback pattern analysis, improving anomaly detection, fraud prevention, and content abuse identification precision by 20% without increasing false-positive review volume.

Deployed and scaled machine learning and LLM models on GCP Vertex AI, leveraging Big Query for large-scale feature aggregation to support recommendation and moderation use cases with 100K+ daily low-latency inference requests.

Operationalized AI and agentic AI solutions using Kubernetes-driven MLOps pipelines, enabling controlled releases, monitoring, and rollback, while delivering dashboards to track model performance, drift, policy violations, and operational efficiency for business stakeholders.

Implemented agentic AI observability (Arize AI / Phoenix-style) using OpenTelemetry tracing, prompt and tool-call telemetry, RAG retrieval metrics, and embedding drift detection, improving production reliability and accelerating root-cause analysis and debugging. Zimmer Biomet USA

Python Developer Jan 2023 – Aug2024

Developed Python and PySpark-based machine learning pipelines to support forecasting and classification use cases for enterprise clients, processing 500K+ records per day and improving model training efficiency by 30%.

Engineered a HIPAA-compliant health scoring platform using Python (FastAPI), PostgreSQL, and React, streamlining patient assessments and reducing manual documentation by 60%, improving triage efficiency and clinical data consistency.

Designed a schema-first GraphQL API layer to unify access to patient records and risk models, reducing payload size by 40% and improving API response times by 28%, enabling near real-time dashboard refresh during peak usage.

Optimized backend performance with asynchronous Fast API endpoints using Python asyncio, reducing p95 latency by 22% and enabling real-time updates for emergency triage workflows.

Achieved 98% automated test coverage by building comprehensive test suites with PyTest, validating REST/GraphQL APIs via Swagger/OpenAPI and Postman, reducing post-release defects by 60% across 2 production launches.

Implemented database migration and versioning using Alembic for PostgreSQL, maintaining schema consistency across dev/stage/prod environments and improving reliability of staged feature rollouts.

Improved PostgreSQL query performance through indexing, query tuning, and execution plan optimization, reducing dashboard load times by 60% for data-intensive clinical workflows.

Automated CI/CD and IaC provisioning using GitHub Actions + Terraform, eliminating environment configuration errors and enabling zero-downtime deployments across multiple staging and production environments.

Integrated AWS Lambda for asynchronous processing (file uploads + medical document parsing), reducing backend compute load by 30% and improving API responsiveness during high-traffic periods.

Architected real-time event streaming and notifications using AWS Kinesis + WebSockets, reducing critical alert delivery for high-priority patient risk events.

Implemented end-to-end observability using OpenTelemetry distributed tracing with AWS CloudWatch (plus Datadog RUM for frontend monitoring), improving root-cause analysis and reducing recovery time during traffic spikes.

Integrated a TensorFlow-based risk prediction model into the backend inference workflow, improving alert accuracy and enabling proactive interventions for high-risk patients.

Integrated NoSQL storage (Amazon DynamoDB) for high-velocity patient events and audit logs, enabling low-latency lookups at scale and improving reliability of real-time alerting workflows. Intex Technologies India

Python Developer July 2019 – May 2021

Built RESTful microservices in FastAPI using Pydantic schemas and versioned endpoints (v1/v2), improving API contract stability and reducing breaking changes across releases.

Implemented structured logging (JSON logs) and request correlation IDs across services, improving production debugging and reducing mean time to resolution (MTTR).

Added rate limiting, request validation, and idempotency keys for critical endpoints, improving API reliability and preventing duplicate processing under retries.

Implemented database transaction management and optimistic concurrency controls to ensure data integrity for high-frequency writes and concurrent updates.

Optimized SQLAlchemy query patterns (bulk operations, eager loading, pagination), reducing query latency and improving throughput for high-traffic endpoints.

Developed background job orchestration with retries, dead-letter handling, and exponential backoff, improving failure recovery and reducing dropped tasks.

Integrated object storage (Amazon S3) with pre-signed URLs for secure uploads/downloads, reducing backend load and improving file transfer performance.

Implemented OAuth2/JWT authentication with RBAC and audit logging to secure PHI access and meet HIPAA compliance requirements.

Implemented feature flags (e.g., LaunchDarkly-style patterns) to enable safe rollouts, A/B testing, and rapid rollback with reduced deployment risk.

Created data ingestion pipelines for batch + streaming inputs, validating and normalizing data before persistence, improving data quality and downstream analytics accuracy.

EDUCATION

Master’s in computer science

Southern Arkansas University, AR, USA

Contact this candidate