Nick Smit

Senior AI / Full Stack Engineer · RAG · RLM · Agentic AI

Senior AI / full-stack engineer (12+ yrs) and lead contributor at Delta Labs — Zurich AI startup, €4.4M seed Mar 2026. I ship production AI systems end-to-end: RAG, Recursive Language Model (RLM) inference, agentic workflows (MCP), OCR pipelines, plus the prompt orchestration and reliability layer (throttling, retry, eval loops) that makes them production-grade. Track record of measurable wins — 3,370× vector-search speedup, 5% → 85% test coverage, 320+ PRs in 10 months. Prior CTO at a healthcare-data startup; acquired exit as 1st technical hire at RTK.io / Magnite.

Experience

  1. Senior Fullstack / AI Engineer (Lead Contributor)

    Delta Labs AG · Zurich, Switzerland

    Lead contributor at Delta Labs — UZH spin-off building Elaiia, an AI customer- simulation platform powered by Recursive Language Models (RLM). €4.4M seed round led by Cusp Capital (Mar 2026). Owning end-to-end delivery across the AI / LLM stack, backend services, data pipelines, and frontend.

    • Built the production AI stack end-to-end on Azure + Postgres + pgvector: RAG retrieval over proprietary enterprise documents (ingestion → OCR → chunking → embedding → indexing), Recursive Language Model (RLM) inference + persona-generation services, prompt orchestration with evaluation loops, throttling, retry / circuit-breaker reliability guardrails.
    • Drove SQL / DB performance: 3,370× vector-search speedup (25s → 7ms on 162k rows) via pg_diskann + index strategy redesign; 10× query speed-ups via materialised views, indexing, schema normalisation; co-designed schema decisions with senior engineers.
    • Owned the testing suite + CI/CD: lifted backend test coverage 5% → 85% (100% on async / event-driven functions, ~8,500 LOC of new tests across 21 files), set code-review and merge-process discipline, hardened GitHub Actions pipelines.
    • Introduced the team's AI Agent Skills + MCP (Model Context Protocol) framework — codified how Claude / Copilot / ChatGPT-style coding agents auto-discover release, ticket, and DB workflows. Standardised AI-assisted productivity team-wide.
    • Architected and shipped a greenfield FastAPI microservice (data-service) replacing the legacy distribution model — significantly more efficient, with weighted distributions, multi-country support and reference validation.
    • Python
    • FastAPI
    • Azure
    • Postgres + pgvector
    • TypeScript
    • Next.js
    • LLM
    • RAG
    • RLM
    • MCP
    • OCR
    • Docker
    • GitHub Actions

  2. CTO

    Ranova · Geneva, Switzerland

    CTO at an early-stage healthcare data startup. Owned architecture, delivery, hiring direction, and product roadmap.

    • Spearheaded migration from a legacy Docker-VM stack to AWS serverless (Lambda, DMS, GitLab CI/CD), reducing infra cost and improving deploy ergonomics.
    • Boosted ETL pipeline throughput ~300% via intelligent orchestration (Inngest + AWS Lambda / Batch).
    • Cut build + deploy time by 55% through CI/CD overhaul; introduced linter and review/merge process to raise code quality bar.
    • Optimised database query speed up to 10x with caching, indexing, and materialised-view strategies; designed NeonDB schema for safer dev workflows and clean prod/staging sync via AWS DMS.
    • Built Flask-based analytics routes integrating LLMs for enhanced insights and exposed them via Streamlit dashboards.
    • Python
    • Flask
    • FastAPI
    • Next.js
    • TypeScript
    • AWS
    • AWS Lambda
    • AWS DMS
    • Inngest
    • NeonDB
    • PostgreSQL
    • Docker
    • GitLab CI/CD

  3. Software Engineer

    EPFL — Blue Brain Project · Geneva, Switzerland

    Built UI and data tooling for EPFL's Blue Brain — a research programme mapping and simulating a mouse brain.

    • Principal developer on neuron-morphology and electrophysiology viewers (React, Next.js) over a 1M+ node neuroscience knowledge graph.
    • Designed the data layer for the front-end app, consuming several sources of complex neuroscience data; reduced query times via Elasticsearch + triple-graph DB strategies.
    • Implemented an LLM-based document generation system that reduced documentation time by ~65%.
    • Deployed microservices via Docker / Kubernetes with accessibility and performance optimisations; ensured responsive, accessible UIs.
    • Collaborated with neuroscientists, designers, backend engineers and data scientists to translate research workflows into usable digital laboratory tools.
    • TypeScript
    • JavaScript
    • React
    • Next.js
    • Elasticsearch
    • Triple Graph DB
    • JSON-LD
    • AntD
    • Tailwind
    • Cypress
    • Python
    • Docker
    • Kubernetes
    • REST

  4. Software Engineer (Principal Frontend / Fullstack)

    RTK.io · Zurich, Switzerland

    Joined as 1st technical hire at an early-stage ad-tech startup; matured the platform to acquisition by Magnite. Principal developer on a SaaS serving 10k+ active users.

    • First tech hire at an early-stage startup that matured and was acquired by Magnite (industry-leading sell-side platform).
    • Principal developer on real-time SaaS ad-tech infrastructure (AngularJS / PHP / Laravel), supporting 10k+ active users.
    • Built a Streamlit-based data monitoring UI that cut anomaly-detection time by 20% and improved overall pipeline reliability.
    • Improved pipeline performance with Redis caching and SQL-query optimisation; led DevOps work on Docker images and GitLab CI/CD.
    • Increased team delivery speed ~20% via agile / Scrum process improvements; led daily stand-ups; contributed to hiring.
    • AngularJS
    • JavaScript
    • PHP
    • Laravel
    • Python
    • Streamlit
    • Redis
    • MySQL
    • Elasticsearch
    • Docker
    • GitLab CI/CD

Selected wins

Selected projects

Skills

AI / LLM
Production RAG architecture, Recursive Language Models (RLM) inference & persona generation, Agentic workflows & MCP (Model Context Protocol), Prompt engineering & orchestration, LLM evaluation loops & metrics, OpenAI / Anthropic / Azure OpenAI APIs, Embedding pipelines (ingestion → chunking → vectorisation → indexing), OCR & document parsing pipelines, Vector DBs (pgvector + DiskANN, FAISS, Pinecone), Throttling, retry, circuit-breaker reliability patterns, Schema-enforced structured outputs
Python
Python 3 (12+ yrs), FastAPI, Flask, ETL pipelines, Pytest, asyncio, Pandas / NumPy
Frontend
TypeScript (8+ yrs), React (8+ yrs), Next.js, Node.js, Tailwind, AntD, Streamlit (4+ yrs)
Data & Storage
PostgreSQL (incl. pgvector + DiskANN), Elasticsearch, Triple / Graph DBs, Redis, NeonDB, MySQL, Knowledge graphs (1M+ nodes)
Cloud & DevOps
AWS (Lambda, Batch, DMS, S3), Azure, Docker, Kubernetes, GitHub Actions, GitLab CI/CD, Vercel, Inngest, ArgoCD
Observability
Structured logging, Metrics & tracing, Coverage reporting, Cost / latency / quality dashboards

Education

  • Northeastern University B.Sc., Criminal Justice / Economics, Boston, USA
  • Exeter University B.Sc. (read), Economics, Exeter, UK
  • Metis Data Science Immersive (ACCET-accredited), New York, USA
  • Dev Bootcamp Full-Stack Web Development, Chicago, USA

Languages

  • English Native / Bilingual
  • French Professional Working (B1+)
  • Afrikaans Professional Working