lachine mearning

Online alias: Ri
March '07
USA|GBR
Education: Computer Science with AI BSc (Hons) (in progress)
Profession: Data Scientist/ML Engineer (contracted)
Extra Curriculars: Lead of AI Society, Course Rep
Goal: Deep Learning Researcher (I'm genuinely going to die with how busy I am)

Fun facts about the larp

I have typographic memory- I can recall text as a literal graphic in my brain. During exam season I'd remember my notes word for word by visualing it
14 piercings and 9 tattoos including logic gate circuits on my arms :3
Anime/manga obsessed: Neon Genesis Evangelion, Psycho Pass, Tokyo Ghoul, Made in Abyss, Attack on Titan, Chainsaw Man, Blame!, Berserk
Severely lactose intolerant but LOVE cheese- could wolf a bag of grated mature cheddar

Larp of all trades and chud of none (joking)

Training models

I train AI models both traditional and deep learning on complex supply chain data across numerous industries to help uncover patterns and risk intelligence for our clients- mixed with LLMs because they're pretty good at semantic reasoning when they're not hallucinating.

tune_xgb.py

import optuna
from xgboost import XGBClassifier

def objective(trial):
    params = {
        "max_depth": trial.suggest_int("max_depth", 3, 10),
        "learning_rate": trial.suggest_float("learning_rate", 0.01, 0.3),
        "subsample": trial.suggest_float("subsample", 0.6, 1.0),
    }
    model = XGBClassifier(**params)
    model.fit(X_train, y_train)
    return model.score(X_val, y_val)

study = optuna.create_study(direction="maximize")
study.optimize(objective, n_trials=50)

[I] Trial 47 finished, value: 0.912
[I] Best params: {'max_depth': 6, 'learning_rate': 0.084}

Data pipelines

I don't just build the models, I deploy them too- sometimes simple statistical techniques rather than ML. Familiar with Google Cloud, Cloud Run, Cloud Scheduler cronjobs, SQL, and Data Studio dashboards.

bash: gcloud deploy

$ gcloud run deploy risk-pipeline \
    --image gcr.io/supply-risk/engine:v2.1

✓ Service deployed successfully.

$ gcloud scheduler jobs create http daily-eval \
    --schedule="0 6 * * 1-5"

Agentic pipelines

I spend more time system designing and prototyping than deploying BECAUSE when I deploy something, you can trust it's ready. Big up LangGraph

agent_flow.py

from langgraph.graph import StateGraph, END

workflow = StateGraph(SupplierState)
workflow.add_node("assess_risk", model_node)
workflow.add_conditional_edges(
    "assess_risk",
    lambda s: "escalate" if s["score"]>0.75 else END
)

Languages

Python, C++, Java, C#, HTML, CSS, JS, SQL

Data science

NumPy, Pandas, SciPy, Matplotlib, Seaborn, Scikit-Learn, PyTorch, XGBoost, OpenCV, StatsModels

Agentic knowledge

LangGraph, LangChain, LLM APIs, DSPy

MLOps

Google Cloud, Cloud Run, BigQuery, Data Studio

Tools

Git, Jupyter

YouTube: iris-lachinemearning · Discord: @spinekisser · GitHub: cat7enthusiast

iris data science