FINN SOMMER / MANNHEIM, DE / 2026 / Available for contract work

I build AI showcases at SAP
and ship the web on the side.

ROLE Working Student, SAP Experience Services
UNIVERSITY M.Sc. Management · U. Mannheim (expected 2027)
B.SC. Business & Computer Science · Grade 1.9
ABROAD U. Wollongong · Grade 1.0
LANGUAGES DE native · EN fluent · FR intermediate
9,000+
attendees at flagship events
4 years
at SAP across 4 roles
MEE
Innovation Award · 2022
  1. 2025 – 2027

    University of Mannheim

    M.Sc. Management · expected

  2. JUL – NOV 2024

    University of Wollongong, Australia

    Study Abroad — Accounting & Computer Science · Grade 1.0

  3. 2021 – 2025

    University of Mannheim

    B.Sc. Business and Computer Science · Grade 1.9

LANGUAGES German native English fluent French intermediate
📚

Studying

M.Sc. Management at the University of Mannheim. Bachelor in Business and Computer Science with a 1.0 study-abroad semester at the University of Wollongong.

Working

Working Student at SAP Experience Services — building AI product showcases for SAP Joule and delivering flagship events from Berlin to Las Vegas.

Building

A growing collection of small web projects: e‑commerce sites, brand experiences, and AI‑powered workflows for friends, founders, and personal experiments.

🌍

Open to

Contract work, internships, and project-based engagements at AI‑native startups. Remote-friendly. Async-first. Output over hours.

Four roles, four years, one company.

From the CTO Office to the Experience Center stage — here are the things I shipped.

0
attendees reached at flagship events
0
visitor engagements at SAP NOW Berlin
0
live viewers per training session
0
MEE Region Innovation Award
  1. APR 2025 NOW

    SAP Experience Services

    Working Student
    • Built customer-ready product showcases for SAP Joule (AI), translating product strategy into compelling enterprise demos.
    • Spearheaded onsite Experience Center delivery at SAP NOW Berlin 2025 (~1,000 attendees) — drove 400+ visitor engagements and generated 8 qualified leads.
    • Executed Experience Center operations at SAP Sapphire Madrid (5,000+ attendees) and SAP Select Las Vegas (3,000+ attendees).
    • Delivered enterprise demos and deep dives on stage — tailored narratives to audience seniority and intent.
    • Coordinated external agencies and partners across logistics, booth deliverables, and demo quality standards.
  2. SEP 2023 FEB 2024

    CTO Board Office — Technology & Innovation

    Working Student
    • Coordinated CTO Executive Board Office operations — scheduling, stakeholder alignment, meeting follow-ups.
    • Prepared executive-ready briefings, agendas, and decision materials for strategic topics and customer meetings.
    • Drove cross-unit project alignment between Executive Board functions and business units.
  3. JAN 2023 AUG 2023

    Customer Advisory

    Working Student
    • Presented sustainability product demos — Sustainability Control Tower, Sustainability Footprint Management, EU taxonomy.
    • Supported core delivery team for Customer Success Summit Germany 2023 at Deutsche Bank Park.
    • Automated monthly leadership reporting via Power Automate — eliminated several hours of manual work each month.
  4. FEB 2022 DEC 2022

    Office of the CTO Germany

    Internship & Working Student
    • Led workstreams for an award-winning internal training series — MEE Region Innovation Award — streamed live to 200+ sales colleagues with hundreds more on replays.
    • Orchestrated onsite customer engagements including board-level meetings and large-scale venues.
    • Owned green-screen studio operations — produced leadership communication recordings end-to-end.
    • Presented first customer showcase at the Greentech Festival in Berlin.

Validating Stylistic Similarity Metrics in Author-Fine-Tuned Large Language Models.

How well do quantitative style metrics actually capture the “voice” of an author once you fine-tune an LLM on their work? I built a full experimental pipeline around LLaMA‑3.2‑1B, fine-tuned it on Charles Dickens and Mark Twain corpora using LoRA, and then evaluated the resulting models against a range of stylometric metrics — from constituency-parsing-based syntactic features to lexical and structural similarity scores.

The result: a clearer picture of which metrics meaningfully reflect author-level stylistic adaptation, and which are noisy or misleading. The repository contains the full training, inference, and analysis pipeline.

View on GitHub

The tools I actually use.

  • Python LANG
  • PyTorch & 🤗 Transformers AI
  • TypeScript WEB
  • n8n AUTOMATION
  • Cloudflare Workers EDGE
  • Adobe Premiere CRAFT

ALSO COMFORTABLE WITH Java · SQL · HTML & CSS · LoRA & PEFT · Claude & GPT APIs · Prompt engineering · Power Automate · Make / Zapier · Lightroom · Photoshop · PowerPoint (executive-grade) · Excel

Working on something interesting?
Let’s talk.

I read every email. Best for: AI product work, web builds, automation projects, or just a good conversation about startups and software.