Charles Cabatian
Manila, PH
I'm a Computer Science student focused on machine learning and deep learning engineering — working across the full pipeline from data modeling to fine-tuning and deployment. I'm especially drawn to research at the intersection of deep learning and edge AI.
Now
Machine Learning Engineering Intern
FlyRank AI
Notion Campus Leader
National University – Manila
Assistant Vice President (Internals)
PyTorch Campus Labs – NU Manila Chapter
Content Creator
Independent
Deep Learning Research · Edge AI · Small Language Models · Retrieval Systems
Things I've Built
cpnlookup 2026 A local-first CLI for natural-language search over GitHub repos.
A local-first CLI for natural-language search over GitHub repos.
A local-first code intelligence CLI that enables natural-language querying over GitHub repositories without requiring a local clone. Uses hybrid retrieval and repository analysis to improve code-context relevance, with support for private local inference.
Python · Hybrid Retrieval · Code Intelligence · Ollama
Nura 2026 A multilingual healthcare navigator for Filipino users, built on RAG.
A multilingual healthcare navigator for Filipino users, built on RAG.
A multilingual healthcare navigation platform designed to help Filipino users navigate healthcare facilities, documentation, and benefits through conversational AI and retrieval-augmented generation.
FastAPI · React · Supabase · RAG
AI-Powered Lead Qualification Pipeline 2026 An autonomous pipeline that researches and ranks business leads.
An autonomous pipeline that researches and ranks business leads.
An autonomous lead qualification system combining web data extraction, retrieval, and LLM-based scoring to research and rank potential business leads.
RAG · LLMs · Web Data · Docker
EIF Store Site Selection Engine 2026 A synthetic data engine for modeling retail site expansion in the Philippines.
A synthetic data engine for modeling retail site expansion in the Philippines.
Built for Eskwelabs' EIF Data Modeling Track: a multi-table synthetic data engine simulating a realistic retail site-selection pipeline across Philippine regional tiers. Parameters are anchored to PSA census density figures and 2025/2026 commercial rent indices, with a deliberately hidden accessibility confounder used to teach analysts to detect second-order effects in candidate-site performance models.
Python · NumPy/Pandas · Synthetic Data · Simulation
E-Commerce Churn Prediction 2026 A model that flags e-commerce customers likely to churn.
A model that flags e-commerce customers likely to churn.
A churn-prediction pipeline for e-commerce customer data — cleaning and feature engineering feeding a classification model used to flag accounts at risk of leaving before they do.
Python · Scikit-learn · Feature Engineering · Classification
Field of Study & Research
Deep learning architecture and applied DL — understanding how things work by rebuilding them.
SLM Supervised Fine-Tuning Study 2026 Does prompt repetition help small language models extract structure?
Does prompt repetition help small language models extract structure?
An empirical study on instruction fine-tuning for small language models — investigating whether the prompt-repetition effect can be internalized through supervised fine-tuning to improve structured information extraction.
PyTorch · Transformers · QLoRA · SFT
Transformer from Scratch 2026 Rebuilding attention and the Transformer block from first principles.
Rebuilding attention and the Transformer block from first principles.
An implementation of core Transformer and attention components from first principles, built to understand how modern language models actually work under the hood rather than through a library abstraction.
PyTorch · Attention · Architecture
RAG from Scratch 2026 An end-to-end retrieval-augmented generation pipeline, hand-built.
An end-to-end retrieval-augmented generation pipeline, hand-built.
A hands-on implementation of an end-to-end retrieval-augmented generation pipeline, built to understand the separation between data, retrieval, and generation layers rather than relying on a framework's defaults.
Python · Retrieval · RAG
WIP Also brewing: an agritech deep learning study.
Experience
JUL 2026 — PRESENT
Machine Learning Engineering Intern
FlyRank AI
Building embedding-based clustering and intent-classification models for search and content-intelligence data.
2026 — PRESENT
Assistant Vice President (Internals)
PyTorch Campus Labs – NU Manila Chapter
Leading chapter operations and partnerships for PyTorch's official campus community at NU Manila.
2026 — PRESENT
Deep Learning Skill Track Scholar
Data Engineering Pilipinas x DataCamp
Currently focused on the Deep Learning skill track as part of the DEP x DataCamp scholarship.
2026 — PRESENT
Content Creator
Independent
Documenting build progress, projects, and lessons learned across ML and software work.
JUN 2026 — AUG 2026
Data Modeling Intern
Eskwelabs (EIF)
Completed the EIF Data Modeling Track — built structured assumptions, synthetic datasets, and decision-support models for real organizational systems.
APR 2026 — JUN 2026
AI & ML Scholar
AWS AI & ML Scholars Program (AWS x Udacity)
Completed the AI Practitioner learning track — generative AI, NLP, computer vision, and responsible AI — with hands-on projects on Amazon Bedrock and PartyRock.
2025 — PRESENT
Notion Campus Leader
National University – Manila
Helping build the Notion community through workshops, events, and productivity-focused initiatives.
Technical Focus
ML & Deep Learning
PyTorch, Scikit-learn, NumPy, Transformers, Hugging Face, Jupyter Notebook
LLM / SLM Systems
RAG Pipelines, Supervised Fine-Tuning, LangChain, Groq API, Google ADK
Backend & APIs
Python, FastAPI, Flask, Node.js
Frontend
TypeScript, JavaScript, React, Next.js
Data & Infrastructure
Supabase, PostgreSQL, Firebase, Docker, Google Cloud Platform, Git
Interests / Rabbit Holes
Deep Learning & Applied Research
Model architectures, fine-tuning, and the parts of a paper that never make it into the abstract.
Edge AI & Quantization
Making models small enough to run somewhere they were never really meant to.
Retrieval & Agentic Systems
RAG pipelines, small language models, and getting agents to actually finish the task.
Hardware & Homelabs
Arduino, OpenCV, self-hosted infrastructure, and whatever breaks next.
Recognition
Certifications
AWS AI & ML Scholars — 2026 Challenge Completion
AWS x Udacity
Deep Learning Skill Track
Data Engineering Pilipinas x DataCamp
Budget AI Part 4: Measuring ROI
Eskwelabs (EIF)
Let's Talk.
If you're working on something interesting, I'd love to hear about it.
Email me → cjcabatian5@gmail.com