Agents

Builds production RAG pipelines - chunking strategies, embedding selection, vector search, reranking, and evaluation for LLM knowledge-retrieval systems.

Decides when and how to fine-tune LLMs - dataset construction, LoRA/QLoRA configuration, training setup, and evaluation for production-adapted models.

Designs and runs LLM evaluation suites - LLM-as-judge, regression testing, benchmark construction, and CI integration to catch AI quality regressions before production.

Selects and configures vector databases - pgvector, Pinecone, Weaviate - with correct index parameters, metadata filtering, and production performance tuning.

Builds production vision pipelines - Claude vision API, image preprocessing, OCR, classification, and structured extraction from images and documents.

Builds production voice pipelines - Whisper transcription, TTS selection, streaming audio, word-level timestamps, and latency-optimized speech AI systems.

Selects and deploys text embedding models - MTEB benchmarks, batching, caching, semantic search, clustering, and production embedding pipeline optimisation.

Deploys and monitors ML models in production - model registry, serving infrastructure, A/B testing, drift detection, and automated retraining pipelines.

Eliminates React performance bottlenecks - re-render diagnosis, memoization, code splitting, virtualization, and React DevTools profiling.