Data Architect and backend engineer with 5+ years of remote experience designing data models and ELT pipelines, building contract-driven APIs, and shipping AI/ML systems for automation, content generation and LLM observability. I like owning a system end to end — schema and storage design, service architecture, cloud deployment and test coverage — with a bias toward reproducibility, auditability and clean interfaces.
| jdsotomarin@gmail.com | |
| GitHub | github.com/josedaniel-dev |
| Phone | +57 316 168 8900 |
| Location | Remote · Colombia (UTC−5) |
- Data architecture & modeling — relational schema design (PostgreSQL, SQLite), migrations with Alembic, ELT/ETL pipeline design, contract- and schema-driven data flows, lineage and auditability.
- Backend & APIs — FastAPI and Flask service design, REST contracts, JWT auth, async Python, and unit + integration testing with pytest.
- AI/ML & LLM systems — RAG pipelines (chunking, vector search), LLM observability and evaluation, per-model cost accounting, TTS/audio generation, and model integration across OpenAI, Anthropic, Google Gemini and Cartesia.
- Cloud & delivery — Google Cloud (Firestore, Secret Manager, Cloud Storage,
Compute Engine, Vertex AI), Docker/Podman, CI/CD, and
systemddeployments. - Data engineering & analysis — Pandas, SQL, scikit-learn, feature engineering, and turning notebooks into reproducible pipelines.
All repositories below are public. Each entry: repository, primary stack, and what was designed or built.
Magic-Lab-Studio/observer · Python + TypeScript · Apache-2.0 · PyPI · npm
Local-first, self-hosted observability and evaluation platform for LLM and agentic systems. Designed the trace data model and async storage layer (SQLAlchemy over SQLite/PostgreSQL with Alembic migrations), multi-framework auto-instrumentation for OpenAI, Anthropic and LangChain, per-model cost accounting across 20+ models, an LLM-as-Judge evaluation engine driven by YAML rubrics plus human feedback, and a real-time dashboard that streams traces over WebSocket.
Stack: FastAPI, async SQLAlchemy, Alembic, Pydantic v2, WebSockets, SQLite/PostgreSQL, Docker, TypeScript SDK
josedaniel-dev/GenAI-Audio-API · Python
Production-grade audio personalization engine built around Cartesia Sonic-3 TTS. Contract-driven API with deterministic stem generation and caching (content signatures, TTL, compatibility checks), a bit-exact semantic merge engine that avoids resampling, a rotational engine for large-scale stem automation, optional Google Cloud Storage consistency auditing, and a 36-test suite spanning generation, assembly, cache, CLI and GCS mock layers.
Stack: FastAPI, Cartesia Sonic-3, Google Cloud Storage, pytest, Makefile-driven CLI
josedaniel-dev/rgmfn · Python
Turns PDF collections into a queryable, semantic knowledge base. PDF merging and text chunking feed a Retrieval-Augmented Generation engine on Google Gemini with a FAISS vector index, plus document chat, audio transcription, ElevenLabs voice answers, Google Drive and Confluence export, and optional Datadog/Sentry/PostHog instrumentation.
Stack: FastAPI, Google Gemini (google-genai), FAISS, ElevenLabs, Google Drive & Confluence APIs, Docker
josedaniel-dev/anyone_ai_sprint3_ml_api · Python
FastAPI service with a decoupled ML worker: JWT auth, user management, feedback persistence and image classification. Relational schema on PostgreSQL, containerised with Docker Compose, and full unit + integration coverage with pytest.
Stack: FastAPI, PostgreSQL, JWT, Docker Compose, pytest
josedaniel-dev/did_pro_provisioning_service-safe_mode · Python
FastAPI microservice that automates acquisition, registration, rotation and health monitoring of Direct Inward Dialing (DID) numbers for high-volume voice campaigns. Twilio integration with a sandboxed SAFE build used as the operational baseline for endpoint and logic validation.
Stack: FastAPI, Twilio, local/JSON storage, pytest
josedaniel-dev/olist-ecommerce-data-pipeline · Python + Jupyter
Extract → Transform → Load pipeline over a LATAM e-commerce dataset. Loads raw CSVs into a relational store, applies a library of parametrised SQL transformations, and produces revenue and delivery-performance metrics for 2016–2018; modelled with SQLAlchemy and covered by pytest.
Stack: Python, SQLAlchemy, SQLite, Pandas, SQL, pytest
josedaniel-dev/airbnb_price_prediciton · Jupyter
End-to-end analysis notebook: ingests Airbnb Open Data, cleans and engineers features, runs exploratory analysis, and trains baseline regression models to predict nightly price.
Stack: Python, Jupyter, Pandas, scikit-learn
josedaniel-dev/scrapper · Python · MIT
Scraper that walks a paginated e-commerce catalogue with requests +
BeautifulSoup, extracts structured product records with resilient parsing, and
exports them to CSV/JSON for downstream cleaning and classification.
Stack: Python, requests, BeautifulSoup, CSV/JSON
josedaniel-dev/TW-C · Python
Defensive-grade execution engine for delta-neutral funding-rate arbitrage that
prioritises capital integrity over latency: idempotent order execution keyed by a
deterministic clientOrderId, decoupled entry and emergency-liquidation circuit
breakers, epsilon-tolerance neutrality checks, and forensic auditing in Firestore
with secrets in Google Secret Manager. Async CCXT connectors, deployed on Google
Compute Engine under systemd.
Stack: Python 3.12, asyncio, CCXT, Google Cloud Firestore, Google Secret Manager, Podman, systemd
josedaniel-dev/focus · Python
Desktop task manager in Kivy: tasks with due dates, periodic reminders and sound effects, local JSON/CSV persistence, and a small embedded Flask REST backend on a background thread. Ships as a monolith plus an in-progress modular refactor.
Stack: Python, Kivy (kvlang), pygame, Flask
| Area | Tools |
|---|---|
| Languages | Python, SQL, Bash, TypeScript / JavaScript, C# |
| Backend & APIs | FastAPI, Flask, REST, WebSockets, Pydantic, asyncio, JWT |
| Data | PostgreSQL, SQLite, SQLAlchemy, Alembic, Pandas, NumPy, FAISS |
| AI / ML | scikit-learn, RAG, LLM evaluation & observability, OpenAI / Anthropic / Gemini / Cartesia APIs, TensorFlow, PyTorch |
| Cloud & DevOps | Google Cloud (Firestore, Secret Manager, Cloud Storage, Compute Engine, Vertex AI), AWS, Docker, Podman, CI/CD, Git |
| Observability | Datadog, Sentry, PostHog, trace/telemetry instrumentation |
| Audio & Media | ffmpeg, pydub, Cartesia, ElevenLabs |
| Other | Kivy, Unity, Blender, Web design (WordPress, Webflow) |
Soft skills: project planning · technical writing · client communication · process optimization · remote collaboration · bilingual EN/ES.
Data Architect — Globant (client: Temenos) | Remote | 2026–Present Data architecture and modeling for Temenos, a core-banking software platform, as part of Globant's delivery organization.
Python Developer & Technical Consultant (Independent) | Remote | 2020–2025 Architected and delivered backend services, ML pipelines and automation systems for international clients — AI-driven content tools, API integrations, reproducible data workflows and production-grade microservices.
Digital Marketing Consultant — Dr. Julián Soto | Remote | 2022–2023 Built automated content pipelines with Python + the ChatGPT API, improving campaign performance and lead quality.
Freelance Data Labeling Specialist — Outlier, Appen, Quorum | 2023–2024 Produced high-precision annotated datasets for multimodal AI training.
Technical Support & Digital Operations — International Clients | 2021–2025 Supported global teams and platforms (Shopify, Nintendo, Affirm, Vivint, Progressive Leasing) with technical troubleshooting and workflow management.
- Fellowship in Machine Learning and AI — Anyone AI
- B.Sc. Audio Engineering — Brunel University, UK
- HNC Creative Media Production — South Thames College, UK
- Level 2 & 3 Sound Engineering — Lambeth College, UK
Certifications: Python & Data Science · IBM AI Fundamentals · Unity VR Developer · Google IT Support & Cybersecurity · Cisco Linux Essentials · SENA Organizational Agility