I build machine learning systems the way high-performing engineering teams do: from first principles, then production. Before reaching for a library, I implement the algorithm by hand — because the ability to reason about a model's failure modes matters more than the ability to call .fit() on it.
My long-term direction is AI Systems Engineering — the discipline sitting at the intersection of modeling, deployment, and governance that defines how frontier labs like OpenAI, Anthropic, DeepMind, and Meta AI actually ship models into the real world. That's the bar I'm building toward.
class Aman:
def __init__(self):
self.role = "ML Engineer, first-principles trained"
self.focus = ["Machine Learning", "Deep Learning", "MLOps", "AI Governance"]
self.philosophy = "Understand it from scratch, then scale it with tools"
self.currently = "Building deployable ML systems, not just notebooks"
self.target_bar = ["Meta", "OpenAI", "Anthropic", "Google DeepMind", "NVIDIA", "SpaceX"]
self.open_to = ["collaboration", "internships", "research discussions"]
def say_hi(self):
return "Let's build something that ships."
me = Aman()
print(me.say_hi())| Layer | Tools |
|---|---|
| Core |    |
| Modeling |   |
| Shipping |   |
| Stage | Focus | Status |
|---|---|---|
| Foundations | ML algorithms & optimization from scratch — no shortcuts | ✅ |
| Applied ML | End-to-end pipelines on real-world, messy data | ✅ |
| Interactive Tooling | Visual, explorable ML systems people can play with | ✅ |
| Deep Learning & Deployment | Neural nets, MLOps, model serving | 🔄 |
| AI Governance & Monitoring | Oversight and control layers for production AI | 🔄 |
| Frontier-Caliber Portfolio | 5–10 polished projects → AI Systems Engineering roles | 🔜 |
🔬 ML-Algorithms-From-Scratch — the foundation everything else here is built on
A ground-up implementation of core ML — no sklearn.fit() shortcuts, every gradient derived and coded by hand.
| Aspect | Details |
|---|---|
| Stack | Python, NumPy |
| Coverage | Linear, Multiple, Polynomial & OLS Regression · Ridge, Lasso, Elastic Net · Logistic Regression, Perceptron, KNN, Linear SVM · Batch / Stochastic / Mini-batch Gradient Descent |
| Repository | amanml2026/ML-Algorithms-From-Scratch |
Why it matters: proves the math is understood, not just imported — the same instinct that separates researchers from library users at top labs.
🎯 ml-decision-surface-visualizer — interactive Streamlit + Plotly tool
| Aspect | Details |
|---|---|
| Stack | Python, Streamlit, Plotly |
| What it does | Renders SGD decision boundaries updating live as hyperparameters change |
| Repository | amanml2026/ml-decision-surface-visualizer |
Why it matters: shows I can build tools people interact with, not just static notebooks.
🚢 titanic-survival-ml — full applied ML pipeline
| Aspect | Details |
|---|---|
| Stack | Python, Pandas, scikit-learn |
| Pipeline | Cleaning → feature engineering (family size, title extraction) → model comparison → cross-validation |
| Repository | amanml2026/titanic-survival-ml |
Why it matters: a disciplined, complete ML workflow on messy real-world data — the unglamorous 80% of the job.
📊 sales-analysis-dashboard — EDA & business storytelling
| Aspect | Details |
|---|---|
| Stack | Python, Pandas, Matplotlib |
| What it does | Translates raw sales data into decisions through exploratory analysis |
| Repository | amanml2026/sales-analysis-dashboard |
Why it matters: data science is a communication discipline as much as a modeling one.
🛡️ GovernorAI- — a control layer for monitoring AI systems
| Aspect | Details |
|---|---|
| Stack | Python |
| Stage | Early-stage exploration into AI governance and oversight tooling |
| Repository | amanml2026/GovernorAI- |
Why it matters: signals direction toward responsible, production-grade AI — the layer frontier labs increasingly hire for.
| Credential | Issuer |
|---|---|
| 🏆 Supervised Machine Learning: Regression and Classification | DeepLearning.AI & Stanford University (Coursera) |
| 🏆 Google Cloud AI | Google Cloud |
| 🏆 Python for Data Science | IBM |
| 🏆 Microsoft Generative AI | Microsoft |
| 🏆 Mathematics for Data Science | SkillUp by Simplilearn |
learning:
- Deep Learning
- Model Deployment
- AI Governance & Monitoring
building:
- Deployable ML systems (not just notebooks)
- Interactive ML tooling
benchmarking_against:
- Meta AI
- OpenAI
- Anthropic
- Google DeepMind
- NVIDIA
- SpaceX
open_to:
- Collaboration
- Internships
- Research discussionsIf you're working on ML systems, deployment, or AI governance tooling — I'd love to talk.
📍 India
"Understand it from scratch, then scale it with tools."