Oscar Fernandez

Co-Founder & CTO at Intake

oscarfdst@proton.me github.com/17tayyy LinkedIn

Download PDF

Experience

Co-Founder & CTO

Apr 2026 – Present

Intake · Lanzadera · Valencia

AI helpdesk for B2B companies: an AI agent that answers their customers on WhatsApp, web chat and email, with a shared inbox and help center. In production with paying customers.

  • Built and operate the backend, AI agent and AWS infrastructure end to end, shipping 240+ production releases in 7 months across 3 services behind a 6,500+ test suite and a PR-gated CI/CD pipeline.
  • Built the agentic support system: a Claude tool-calling loop over each company's knowledge base (RAG) and its own APIs, with guardrails and human handoff, answering customers on WhatsApp, web chat and email.
  • Designed the evaluation pipeline that gates every AI change (LLM-as-judge on real conversations, retrieval benchmarks, blind model comparisons), used to choose, tune and migrate models.
  • Implemented the Model Context Protocol on both sides: an MCP server with its own OAuth 2.1 authorization server, and an MCP client that connects customers' own tools to the agent.
  • Built channel and business integrations as a separate service: WhatsApp Business as a verified Meta Tech Provider, Slack, email, Shopify, HubSpot, and public APIs for Zapier and Make.
  • Brought the live AWS estate (ECS Fargate, RDS, ElastiCache, CloudFront) under Terraform by import with zero recreated resources, and added monitoring, alerting and backups.
  • Replaced Celery in production with ArdiQ, my own Rust-core task queue, which now runs every background job.
  • Run recurring security audits covering multi-tenant isolation, access control, container hardening and PII scrubbing.

Python, FastAPI, MongoDB (Atlas Vector Search), PostgreSQL, Redis, ArdiQ, Claude (Anthropic API, Amazon Bedrock), MCP, AWS (ECS Fargate, ALB, RDS, CloudFront, ElastiCache, Route 53, Secrets Manager, CloudWatch), Terraform, Docker, GitHub Actions, Sentry

Full Stack Developer Intern

Sep 2025 – Apr 2026

Moodest · Mataró

B2B SaaS AI platform for HR: intelligent helpdesk with conversational agent and people analytics.

  • Developed a FastAPI REST API integrating with a time-tracking platform of 300,000+ users, covering authentication, role management and KPIs.
  • Built 4 channel integrations for the conversational agent, including Slack and Microsoft Teams, from protocol design to deployment.
  • Led development of the people analytics dashboard (Angular, TypeScript, NestJS) and managed Docker and Nginx production deployments on a self-hosted server.

TypeScript, Angular, NestJS, Python, FastAPI, MongoDB, Redis, Docker, Nginx

Projects

ArdiQ

GitHub

Open-source task queue for Python with a Rust core that runs the worker loop and all Redis I/O off the GIL; published on PyPI (v1.2, MIT) and running in production at Intake. Leads CPU-bound and dispatch throughput in a public, reproducible benchmark against Celery, ARQ, Dramatiq, Taskiq and Streaq, enqueuing 81,000 tasks/s on 33 MB of memory.

Rust, Python, PyO3, Tokio, Redis Streams

wong

GitHub

Open-source memory layer for coding agents (Claude Code, Codex, Cursor): one brain per project synced across machines through a self-hosted server, with every note verified against the files, commits and issues it cites. Ships as a single Go binary with a dashboard, an MCP server and optional end-to-end encryption.

Go, SQLite, MCP

Uatu

Dark-vessel detection on the Spanish coast: pairs Sentinel-1 radar detections with recorded AIS to find ships that are not broadcasting, running 24/7 and tracking about 1,300 vessels an hour.

Python, FastAPI, PostGIS, NumPy, Sentinel-1 SAR

Zona 14

zona14.app

AI-powered pre-match football analytics. Built a Dixon-Coles bivariate Poisson goal model, fit per league from football APIs (time-decay form weighting, shrinkage priors, xG targets and closing-odds signals; L-BFGS-B with analytic gradients) and backtested against sharp-bookmaker baselines via closing line value. Its score grid drives betting markets flagged for value, with a grounded LLM layer explaining each pick in plain language.

Python, NumPy, SciPy, Next.js

Security research and LLM benchmarks

Blog

Found and responsibly disclosed an IDOR chain leading to full account takeover in the internship platform used by every vocational student in Catalonia (200,000+ students); patched. Benchmarked Claude Haiku 5.5 over 960 support conversations, showing that disabling thinking made the agent miss 25% of required human handoffs versus 0% at low effort.

Web security, Python

Technical skills

Languages Python, Rust, Go, TypeScript/JavaScript, SQL, Bash

Backend and AI FastAPI, asyncio, SQLAlchemy, Pydantic, NestJS, agentic systems, RAG, LLM evaluation, MCP, Anthropic API, Amazon Bedrock

Data and Infrastructure MongoDB Atlas (Vector Search), PostgreSQL, Redis, AWS, Terraform, Docker, GitHub Actions

Security Web pentesting, multi-tenant isolation, security audits, OSINT, Hack The Box (25+ machines)

Education

Higher Vocational Diploma in Multiplatform Application Development (DAM)

Expected 2028

Prometeo FP by thePower (online), alongside a Cybersecurity Master's

Intermediate Vocational Diploma in Computer Systems and Networks (SMR)

2026

Escola Pia Santa Anna, Mataró

Languages

Spanish Native

Catalan Native

English Professional working proficiency

Additional training

Hack4u (Cybersecurity Academy)

2024

Offensive Python, Introduction to Hacking, Introduction to Linux.