Enterprise AI for 32M+ passengers. Rebuilding ZRH Insider. Voice AI POC to tender in two weeks.
OpenAI · Forward Deployed Engineer (Zurich)
Great pilots die in legal review. I'm the person who gets them live.
Head of AI at Zurich Airport. I write Python and TypeScript, run evals before go-live, unblock nDSG reviews. This page: who I am, one thing you can play with, and patterns you can steal for your next Zurich account.
0+Production AI tools shipped
Python and TypeScript in production.
0M+Passengers in operational scope
Scale context, not a claim I built AI for every passenger.
0+Enterprise clients at KI Unlocked
40% workflow reduction in deployments.
- PwCTransfer pricing
- Credit SuisseChairman's Office
- Zurich AirportAI Lead
- Forbes30 Under 30 DACH
- TEDx2x speaker
- HSGMaster's degree
PwC to Credit Suisse to two Head of AI roles to Zurich Airport
Builder since 19. I ship fast and sell into enterprise.
AI workflow platform. 40% workflow reduction, 8+ enterprise clients.
Repositioned an SAP consultancy as AI-ready with shipped integrations.
Streamlined AI strategy and eliminated delivery inefficiency.
Strategic initiatives and executive support at the highest level.
International tax consulting. First enterprise consulting foundation.
DeFi lending startup. Raised 1,200 ETH, $5.6M funding round.
Harvard co-project on innovative business research.
Forbes 30 Under 30DACH, 2022
2x TEDx SpeakerHEC Paris, EDHEC
19+ winsHackZurich, Bain FinTech, venture.ch
30+ talksEnterprise and startup stages
Leader of TomorrowSt. Gallen Symposium, 2021
HSG Master'sUniversity of St. Gallen
90 seconds. Play it.
Who I am and why I fit this role. No deck.
90 seconds
Why I am the right fit for OpenAI
A direct pitch for the Forward Deployed Engineer (Zurich) role in Zurich.
Video chapters
Eval workbench: de-risking a first production deploymentOpenAI
How I structure eval gates, latency budgets, and compliance checks for a Swiss bank's first OpenAI workload. Adjust parameters to see where a pilot passes demo stage but fails production review.
OpenAI Zurich FDE Eval & Deployment Workbench
LIVE INTERACTIVESimulating production model trade-offs & Swiss nDSG data constraints
Swiss nDSG PII Sanitizer
In-memory redaction & data residency
Structured Output Fallback Chain
Auto-retry schema violations
Temperature0.2
RAG Vector Context Chunk Size4 KB
Real-Time Simulated Performance Metrics Live
Latency (E2E)
224 ms
TTFT ~90ms
Faithfulness Eval
99.5%
Automated benchmark
Cost / 1k Requests
$0.96
Prompt + Completion
Swiss Privacy
100% nDSG
In-Memory PII Mask
PIPELINE EXECUTION TRACE
[0ms] Init payload for gpt-4o (temp=0.2)
[+12ms] Swiss nDSG PII Redactor: Masked 2 sensitive fields
[+90ms] First Token Stream Received (TTFT)
[+214ms] JSON Schema validation passed
[+224ms] Completed with 0 hallucinations | Faithfulness: 99.5%
"Show don't tell: I build systems where performance and compliance go hand in hand." : Terence La
Mapped to what OpenAI FDE hires for
Three things I actually do on FDE accounts.
End-to-end implementation
FDE means living inside the customer's architecture. I debug integrations, tune prompts under load, and ship fixes while ops teams are watching at Zurich Airport.
Eval workbench discipline
Enterprise motion depends on measurable quality gates before rollout. I built post-interaction eval loops that compare model output against operational ground truth.
Regulated account credibility
Swiss banks and infrastructure operators share blockers I solve weekly: nDSG data residency, audit trails, and legal sign-off on token paths.
How I unblock DACH production deployments
The patterns I use when a Swiss enterprise account is stuck between a successful POC and signed production.
Eval gates before go-live
Define precision, faithfulness, and latency thresholds tied to the customer's actual workload, not generic benchmarks.
My proof
Built post-interaction eval loops at Zurich Airport that compare model output against operational ground truth before any feature ships.
Projects with real outcomes
I spot trends early, build things fast. Side ventures that prove execution depth beyond one employer context.
Project intelligence
Project A12DecisionsExpertsStandards
Why did we change the load assumption?3 cited sources
AI workflow platformActive
KI Unlocked
Enterprise AI workflows with measurable ROI. 40% workflow reduction across client deployments.
8+ enterprise clients
MarketAccountsCampaign
From signal to actionWeb3 charityVenture
perseedU
Transparent donation infrastructure with personal tokens. Top 10 at venture.ch in finance.
1,100+ community members
ZRHHKG
Next: Gate E4612 min walk · passport control ahead
SustainabilityHackZurich 2021
ZooMania
Gamified eco-purchases for families. Winner at Europe's largest hackathon.
HackZurich winner
E-commerceExit
NeverAlone.ch
Emergency marketplace for small businesses during COVID. Built and exited in weeks.
50 vendors onboarded in 4 weeks
AI
Procurement negotiationClarity 8.4Control 7.8
Sales & GTMAgency
TL Innovations
Sales funnels, copywriting, and outbound for Swiss enterprise clients including HSG Family Institute.
$0.5M revenue, 10 clients
✓
Context protectedIdentity hidden OnRaw data stored Off
DeFi lendingFunded
SmartCredit.io
Decentralized lending platform. CMO role leading to major funding milestone.
Raised 1,200 ETH