Anna FP
Data AI Engineer — Building intelligent, data-driven systems.
Competencies
AI & LLM Engineering
Building intelligent agents, agentic workflows, and RAG systems with state-of-the-art LLMs.
LLM Observability & Resilience
Ensuring reliable, production-grade LLM applications through rigorous evaluation, tracing, and monitoring.
Machine Learning & MLOps
Designing, training, and deploying ML models with robust experiment tracking and CI/CD pipelines.
Data Engineering
Structuring scalable data pipelines, ETL flows, and vector databases for semantic search.
Backend & APIs
Developing high-performance, asynchronous REST APIs and backend architectures.
Cloud & Infrastructure
Deploying secure, containerized environments and serverless resources in the cloud.
Projects
LLM Lens
Agentic AI Auditor for LLM Visibility & Discoverability
An interactive multi-agent system that audits how visible and machine-readable your website is for AI agents and LLM search crawlers — then drafts the exact files to compliance.
ArtGuide
AI Art Detector and Audio Guide
ArtGuide is an AI-driven solution that recognizes artworks instantly and produces smooth, natural audio descriptions to enhance your experience in museums and exhibitions.
ML Resilience Lab
Resilient Real-Time Data Ingestion & Fraud Detection Pipeline
An experimental playground showcasing resilience patterns, fault tolerance, and chaos engineering principles within production-grade machine learning pipelines.
The News Hub
Real-Time News Engine and Insights
The News Hub is a comprehensive news aggregation and analysis platform driven by AI. It serves as a centralized system to collect, process, and interactively explore global news content efficiently. By combining advanced data engineering with natural language processing, it empowers users to stay informed through structured insights rather than just raw headlines.
Experience
AI Engineer
- •Designed production agentic systems on LangGraph with LLM-powered supervisors, specialized subagents, ReAct reasoning, conversational human-in-the-loop gates, and multi-turn artifact refinement — architecting each component to earn its complexity before adding it.
- •Built and operated LLM evaluation suites (LangSmith + RAGAS + LLM-as-judge) across 20+ test cases with prompt versioning tracking score progression — achieving 85%+ pass rate, gated on CI/CD via GitHub Actions on every push.
- •Designed RAG-backed memory architectures combining vector search (Qdrant, ChromaDB) with hybrid retrieval strategies — local embeddings as primary path, frontier model reserved exclusively for low-confidence edge cases — optimising accuracy against API cost.
- •Engineered production-grade ML pipeline resilience: state-driven Circuit Breakers for zero-timeout API fallback, autonomous Kill Switch triggered by real-time data drift monitoring at 80% distribution deviation, and hybrid LLM/deterministic generation where rule-based logic overrides model output at critical decision points.
- •Validated ML models and agentic pipelines with structured MLOps tooling — MLflow experiment tracking, model registry and REST scoring server, CI/CD-gated eval pipelines — achieving AUPRC 0.931 on a 284K-record imbalanced fraud dataset.
- •Built and self-hosted multimodal AI serving infrastructure (CLIP embeddings, Piper TTS across 3 languages) via FastAPI async APIs, eliminating third-party dependencies and maintaining full control over inference latency and cost.
Data Engineer
@ BMAT Music Innovators- •Engineered robust Python scripts to synchronize and maintain critical backend data flows, guaranteeing continuous data consistency across Dev, Staging, and Prod environments.
- •Developed high-concurrency RESTful APIs using FastAPI and asynchronous programming, providing stable support for multiple downstream services and ensuring strict data validation with Pydantic.
- •Optimized search performance and metadata discovery by implementing and managing Elasticsearch clusters for high-volume music catalog indexing, ensuring millisecond-latency retrieval.
- •Designed and deployed interactive production dashboards, centralizing scattered data to facilitate real-time data visualization and decision-making for key stakeholders.