EANGELICA ATON
TECH INNOVATION
Product Engineer and Precision Health Expert building full-stack AI systems that ship to production.
> 8+ Years Experience
> AI/LLM Specialist
> NYC-Based
01. ABOUT
I am a Product Engineer and Global Senior Product Manager who thrives at the intersection of engineering, product thinking, and AI. I don't just write code; I own features end-to-end—from architectural design to production launch—focusing relentlessly on customer experience and business impact.
With over a decade of experience across the U.S. and Europe, I specialize in building AI-driven frontend experiences and foundational data platforms. My background spans scaling global health data systems serving over 2M patient records, leading LLM-based products that increased retention by 25%, and architecting API frameworks that improved reliability by 40%.
I excel in fast-paced environments where rapid iteration and outcome-driven development are paramount. Whether it's optimizing LLM latency or designing intuitive human-in-the-loop workflows, my goal is always to ship meaningful outcomes, not just features.
AI Engineering
LLMs, Computer Vision, RAG pipelines, and human-in-the-loop systems.
Production Reliability
Observability, monitoring, testing, and risk mitigation for critical systems.
02. FEATURED_WORK
AI Workflow Platform
Production AI system for healthcare workflows with human-in-the-loop validation. Reduced manual effort by 25% while maintaining clinical accuracy.
HACKATHONS_&_INNOVATION
./GIT_HIGHLIGHTS
Demonstrates how LLMs live inside product workflows, not just notebooks. Focuses on abstraction boundaries, human-in-the-loop logic, and provider-agnostic design.
def run_workflow(user_input: str) -> dict:
llm_result = call_llm(user_input)
if requires_human_review(llm_result["confidence"]):
return {
"status": "needs_review",
"result": llm_result
}
return {
"status": "approved",
"result": llm_result
}Clear Abstraction
Swap providers without touching product logic
Human-in-the-Loop
Confidence-based review triggers
Demonstrates production thinking around AI risk with validation, fallbacks, monitoring, and guardrails.
def run_pipeline(input_data: dict):
if not validate_input(input_data):
return fallback_response("invalid_input")
if not enforce_guardrails(confidence, output):
return fallback_response("guardrail_violation")
return {"status": "success", "output": output}Guardrails
Enforce safety checks on outputs
Fallbacks
Graceful degradation on failure
ictus
C++ implementation of a beat/pulsation algorithm, referencing medical definitions of stroke/seizure.
C++pytorch
Forked contribution: Tensors and Dynamic neural networks in Python with strong GPU acceleration.
C++ / PythonspaCy
Forked contribution: Industrial-strength Natural Language Processing (NLP) with Python and Cython.
Python04. TECH_STACK
BACKEND
- > Python (FastAPI, Flask)
- > Node.js / TypeScript
- > C++ / Java
- > SQL / BigQuery
AI / ML
- > PyTorch / TensorFlow
- > LLMs & Transformers
- > Computer Vision
- > RAG Pipelines
FRONTEND
- > React / Next.js
- > TypeScript
- > Tailwind CSS
- > Data Visualization
INFRASTRUCTURE
- > Google Cloud Platform
- > Docker / Kubernetes
- > CI/CD Pipelines
- > Monitoring / Observability
INITIAL0]R,G&X#Q84#>8
Interested in AI-powered products where engineering quality and customer experience both matter? Let's talk.