Project Write-ups

Does the model learn the task, or the hole in the grader?

TRACE: reward hacking in RL fine-tuning

A controlled study of reward hacking: SmolLM2-360M trained with SFT then GRPO against a weak and a strong verifier, with logit lens and probes to find where the two models...

Can five small LLM calls write a puzzle that actually teaches a word?

Vocabscapes: multi-agent puzzle generation

A five-agent word-puzzle pipeline (writer, judge, improver, finalizer, explanation) with bounded retries and cached variants per level.

Can you trust the judge grading your puzzle generator?

An LLM eval harness for puzzle generation

LLM-as-judge harness across four model pairs, calibrated against blind human ratings.

How does a feed learn what you like?

Social media recommendation engine

A TikTok-style recommender built from scratch: scoring, Gemini embeddings, FAISS indexing and the exploitation-exploration trade-off.

Can a model learn from data it never sees?

Privacy-preserving fraud detection with federated learning

Undergraduate dissertation: federated training across 3 clients on Ethereum transactions with 90%+ class imbalance, reaching 99.7% AUC-ROC.

What predicts strong public support for climate action, and does it split the world into types?

UNDP climate survey analysis

Predicting and segmenting support for climate action across 73,000+ respondents in 73 countries: 89.76% AUC-ROC with a random forest, six country typologies with k-means.

All Posts

The ML Fundamentals Handbook

Jan 2026 · 15 min

Fundamental ML concepts with examples and explanations that feel like a conversation, not a lecture. A guide for building your machine learning foundation before you get to know about different...

AWS Cloud Quest Task 6: Connecting VPCs

Jan 2026 · 9 min

Learning how VPC Peering lets different departments communicate directly without submitting tickets every time they need to access shared resources. Secure, private, and way more efficient.

AWS Cloud Quest Task 5: VPCs

Jan 2026 · 9 min

Helping a bank fix their connectivity issues by setting up VPC Internet Gateways. Turns out, VPCs don't come with internet access by default - you gotta configure it yourself.

AWS Cloud Quest Task 4: AWS Pricing (Spoiler: It Adds Up Fast)

Jan 2026 · 4 min

Using AWS Price Calculator to estimate costs for a surf shop's cloud migration. Played around with instance types, scaling, and discovered that cloud bills can get wild real quick

AWS Cloud Quest Task 3: Picking The Right EC2 Instance (it matters)

Jan 2026 · 6 min

Learning about AWS Systems Manager, using EC2 Instance Connect to access servers without SSH keys, and optimizing instance types for better performance

AWS Cloud Quest Task 2: EC2 - When You Need a Server Yesterday

Jan 2026 · 8 min

Learning about EC2 instances, AWS regions, availability zones, and spinning up virtual servers in the cloud. Also featuring my first DIY fail.

AWS Cloud Quest Task 1: Why S3 Is Actually Pretty Cool

Jan 2026 · 8 min

My first impressions of an AWS learning game that feels more like a town-builder than a technical course. Follow along as I create a questionable avatar, dodge a rogue gorilla,...

Making Regression Actually Make Sense: Building an Interactive Visualizer

Nov 2025 · 15 min

An interactive tool for seeing regression: tweak slope, intercept, polynomial order and learning rate, and watch L1 and L2 fits update in real time.