lk KINNUNEN AIRML Flight Operations
Boarding open · available for new roles Boarding pass · please keep this document with you
Boarding Pass · Kinnunen Air first class
Lauri Kinnunen
Name of passenger

Kinnunen Lauri

Machine Learning Engineer ml-eng

I design, train and ship machine learning systems that survive contact with production, from forecasting models that steer real logistics decisions to retrieval pipelines and the MLOps runways that keep them reliable.

Check in View flights
From
tku
Turku, Finland
To
prd
Production, Worldwide
FlightLK 2026
GateML
Seat1A
BoardingNow
Experience7+ yrs
ZoneEET · Remote
Gate A2 · Passenger manifest

Passenger notes

status: on time
Passenger Informationdoc a2

I'm a Machine Learning Engineer based in Turku, Finland, with seven years of experience taking models from notebook prototypes to dependable services. My work sits where modelling meets software engineering: I care as much about data contracts, latency budgets and monitoring as I do about loss curves.

I've built demand-forecasting systems for logistics, anomaly detection for industrial sensor fleets, and LLM-powered retrieval tools for knowledge-heavy teams. I like small, well-instrumented systems, honest offline evaluation, and the moment a model starts paying for itself.

Outside work I mentor junior engineers, contribute to open-source ML tooling and give talks at local meetups about practical MLOps.

0Years in flight
0Models in prod
0% cost saved
0Talks given
  • End-to-end ownership, from data to deployment
  • Measurable business impact over leaderboard wins
  • Reproducible pipelines and readable code
Gate B1 · Flight log

Legs flown

4 legs · 7 yrs
Leg 04 · Current flightin the air
2023
Departed
now
Arrival

Senior Machine Learning EngineerSaaristo Logistics Analytics · Turku

  • Lead a team of four building forecasting and route-optimisation models used across 140 distribution sites.
  • Replaced legacy statistical forecasts with gradient-boosted and transformer models, cutting forecast error by 21%.
  • Designed the internal ML platform: feature store, model registry, canary deployments and drift alerts.
Leg 03landed
2021
Departed
2023
Arrived

Machine Learning EngineerKide Industrial Systems · Remote

  • Built streaming anomaly detection for 9,000+ industrial sensors with Kafka, PyTorch and ONNX Runtime.
  • Quantised and deployed models to ARM edge gateways, reducing cloud inference costs by 38%.
  • Introduced experiment tracking and automated retraining that cut model refresh time from weeks to hours.
Leg 02landed
2019
Departed
2021
Arrived

Data ScientistMerilinna Retail Group · Turku

  • Developed customer churn and recommendation models serving 1.2M loyalty members.
  • Moved ad-hoc notebooks into tested Python packages and scheduled Airflow pipelines.
  • Ran A/B tests on personalised offers that lifted campaign conversion by 14%.
Leg 01 · Originlanded
2014
Departed
2019
Arrived

M.Sc. Computer Science, Machine LearningUniversity studies · Turku

  • Thesis on probabilistic time-series forecasting with deep state-space models.
  • Research assistant in an applied-AI lab working on medical signal classification.
Gate B2 · Baggage claim

Checked baggage

4 bags · all tagged
Baggage Manifest · Skills on boardweight = proficiency
mdl

Modelling

  • PyTorch23 kg
  • Time-series forecasting22 kg
  • LLMs & RAG19 kg
  • Computer vision16 kg
ops

MLOps

  • Docker & Kubernetes21 kg
  • MLflow & model registry22 kg
  • CI/CD for ML20 kg
  • Monitoring & drift19 kg
dat

Data

  • SQL & dbt22 kg
  • Spark & Polars19 kg
  • Kafka streaming18 kg
  • Airflow / Dagster20 kg
eng

Engineering

  • Python23 kg
  • FastAPI services21 kg
  • AWS & GCP19 kg
  • Rust & C++ (inference)14 kg
scikit-learnXGBoostHugging Face ONNXRayTerraform Weights & BiasesPostgreSQLpgvector PrometheusGrafana
Gate C1 · Connecting flights

Selected flights

departures: 4
Flight LK 101on time
Destination

Northstar: multi-horizon demand forecasting

A forecasting platform predicting daily volumes for 140 warehouses, 28 days ahead. Temporal Fusion Transformer with quantile outputs feeds staffing and inventory planning.

PyTorchTFTDagsterBigQuery
MAPE−21%
Horizon28 d
Sites140
Flight LK 202on time
Destination

Sentinel: edge anomaly detection

Streaming autoencoders flag faults in rotating machinery from vibration data. Models are quantised to INT8 and run on low-power gateways with sub-10 ms latency.

KafkaONNX RuntimeRustGrafana
Latency8 ms
Sensors9k+
Recall0.94
Flight LK 303boarding
Destination

Atlas: retrieval-augmented knowledge assistant

An internal assistant answering questions over 60k technical documents, with hybrid search, re-ranking, cited answers and an evaluation harness tracking faithfulness per release.

LLMspgvectorFastAPIRagas
Docs60k
Hit@50.91
p951.4 s
Flight LK 404on time
Destination

Runway: self-service ML platform

A paved road for data scientists: project templates, a shared feature store, one-command training jobs on Kubernetes and automated canary rollouts with drift monitoring.

KubernetesMLflowFeastTerraform
Deploy15 min
Users35
Models24
Gate C2 · Check-in desk

Check in with Lauri

desk open
Contact Pass · Check-indesk c2
From
you
Your inbox
To
tku
Turku, Finland
First nameLauri
SurnameKinnunen
AddressPiinokankatu 3, 20610 Turku, Finland
Phone+1 814 350 0400
Emailpleasantsrobayocumuto16@gmail.com
Send email Call