Production LLM systems, agents & infrastructure.

I build production software with AI agents — and the LLM evals, observability, and infrastructure that keep it honest. 17+ years in SaaS (AWS, TypeScript). Currently at Aircall, where my systems process 1M+ call transcriptions a day. 9× AWS certified. Former CTO / technical co-founder.

• Agentic systems — tool-using assistants, orchestration, and the backend primitives that make agents safe in production.
• LLM evals & observability — treating model quality as CI (versioned benchmarks, hard gates), and tracing stochastic systems before reaching for dashboards.
• AI infrastructure — event-driven, serverless architectures on AWS (Lambda, Fargate, DynamoDB, Step Functions, CDK); RAG on Bedrock.
• Verifiable, not vibes — the code is open source, the numbers are real, and the essays are published.

• Kaiord · kaiord.com — open-source, local-first training platform. 9 npm packages, a CLI, and an MCP server (in the official MCP registry), plus an evals harness (22 benchmarks behind a 90% CI gate). Every line written by AI agents.
• Deepgent — multi-tenant AI chatbot SaaS: RAG on AWS Bedrock, embeddable widget, fully serverless. Live in production.
• streaming-lambda-ai-sdk — structured LLM responses from Amazon Bedrock (Claude) streamed through AWS Lambda.
• tanita-to-garmin-cdk — serverless AWS app (CDK + Step Functions) syncing body-composition data into Garmin Connect.
• trainingpeaks-sdk — TypeScript SDK for the TrainingPeaks platform.

• Essays → — evals as CI, observability for LLM pipelines, shipping software with AI agents, and GEO. Every claim anchored to public, verifiable artifacts.
• Talks → — available for technical talks on production AI, evals, and agentic development. English & Spanish, Madrid-based.

• Languages: TypeScript / Node.js, Python, Go.
• AWS: Lambda, Fargate, DynamoDB, API Gateway, EventBridge, Step Functions, Bedrock, CDK — 9× certified (incl. Generative AI Developer – Professional, Solutions Architect – Professional, ML Engineer – Associate).
• AI / LLM: agentic workflows, RAG, LLM evaluation, observability, MCP.
• Architecture: event-driven, serverless, hexagonal architecture, domain modeling.

• Aircall (Mar 2025 – Present) — Senior Software Engineer & AI Architect. AI call-summarization, LLM evals & observability, high-throughput serverless pipelines.
• hireful (Apr 2022 – Mar 2025) — Lead Cloud Software Engineer. Migrated legacy systems to cloud-native, event-driven architectures; CI/CD; B2B SaaS delivery.
• Earlier — Software Architect / Technical Lead / CTO / technical co-founder across web, mobile, and SaaS. Full CV → pabloalbaladejo.com/cv.

Triathlete when I'm not shipping.