Skip to content
Quezon City, PHOpen to remote / international

AI/ML engineer,data scientist,backend architect

I work the whole path from raw data to running system: ML pipelines that clean, transform, and analyze datasets into executive BI dashboards, GPU-optimized inference, RAG architectures with real vector isolation, and the async backends that keep them online. Certified DataCamp Data Scientist, reading Computer Science (Data Science). Former FlyRank AI machine learning intern (Jul–Sep 2026), currently co-founder of Dzuka Agri.

  • DataCamp Data Scientist
  • BS CS — Data Science
  • SM Foundation Scholar
Selected workandreipamesa20@gmail.com
Fig. 01 — embedding spacesubject-scoped retrieval
DIM 1DIM 2PHYSICSCALCULUSSYSTEMS
Scopek=5 · 17 vectors in scope

Vectors are partitioned by subject before they are ever searched. The query resolves to its nearest neighbours inside one scope only, which is how cross-subject context contamination stops being a retrieval problem.

Measured impact

  • 0%+

    VRAM cut, AMD MI300X pipeline

  • 0%+

    Retrieval latency reduction

  • 0%

    Structurally validated deck output

  • 0/7

    Andy's Hub production availability

01

Selected work

Each build is stated the same way: the problem that made it necessary, what was actually engineered, and the measured result. Open a card for the full record.

  • 01Problem
    High VRAM overhead and inference latency bottleneck PyTorch models on enterprise GPU clusters.
    02Built
    A 7-phase model compression platform (knowledge distillation, pruning, INT8/FP8 quantization) tuned for AMD Instinct MI300X via ROCm, with an Optuna-powered Pareto frontier engine and an async FastAPI/aiosqlite/WebSockets backend for live benchmarking.
    03Impact
    60%+ VRAM reduction, 2× QPS throughput for high-concurrency production deployments.
    ForgeAI · AMD Developer Cloud Hackathon

02

Measured impact

Numbers reported per project, on the scale they were actually measured on. The throughput figure is a multiplier, so it is plotted separately rather than flattened onto the percentage axis.

Fig. 02 — reported per projectscale: 0–100%

  • ForgeAI

    VRAM reduction

    0%

  • Andy's Hub

    retrieval latency cut

    0%

  • Sophy

    structurally valid output

    0%

0255075100
Fig. 03 — separate scalemultiplier, not a percentage

ForgeAI

QPS throughput vs. baseline

baseline
1.0×
compressed
2.0×

0.0×QPS throughput

03

Experience

Machine learning in production, backend architecture from scratch, and the infrastructure work that keeps a student chapter running.

  1. Jul 2026 — Sep 2026

    Machine Learning Intern

    FlyRank AI

    Engineered automated ML pipelines to clean, transform, and analyze datasets, streamlining data workflows and powering executive BI dashboards.

  2. 2026 — present

    Co-Founder & Lead Backend Engineer

    Dzuka Agri

    Backend architecture, REST API design, and geospatial telemetry ingestion for a multi-agent agricultural diagnostics platform serving smallholder farmers.

  3. Aug 2025 — present

    Junior Officer

    FEU Tech ACM Student Chapter

    Manage chapter technical infrastructure; organize developer workshops and hackathon bootcamps.

04

Stack

What I reach for, grouped by where it sits in the system.

01

Languages

  • Python 3.11+
  • SQL
  • PL/SQL
  • C++
  • Java
  • PHP
  • JSON
02

Data & analytics

  • SQL / PL-SQL
  • ETL / ELT pipelines
  • Data cleaning & transformation
  • Exploratory analysis
  • scikit-learn
  • Embedding & similarity scoring
  • Executive BI dashboards
  • PostgreSQL / SQLite
03

AI / ML

  • PyTorch
  • ROCm workloads
  • Quantization & pruning (INT8/FP8/INT4)
  • Knowledge distillation
  • RAG
  • ChromaDB
  • Multi-agent LLM orchestration (Groq/Llama 3, NVIDIA NIM/Qwen)
  • OCR
04

Backend & cloud

  • FastAPI
  • Pydantic v2
  • REST / WebSockets
  • Supabase (PostgreSQL)
  • SQLite / aiosqlite
  • GCP VMs
  • Cloudflare Tunnels
  • Docker
  • CI/CD
05

Frontend

  • Next.js
  • Tailwind CSS
  • Streamlit
05

Education & certifications

Coursework in data science, plus the certifications backing the applied side.

Education

BS Computer Science — Data Science

FEU Institute of Technology, Manila

Expected Jul 2028100% SM Foundation Scholar

Certifications

  • Data ScientistDataCampAug 2026
  • AWS Certified AI PractitionerAWS2026
  • CCNA: Introduction to NetworksCiscoJul 2026
  • IT Specialist — PythonCertiportValid to Jul 2031