samouzegar

SadafAmouzegar

she/her · AI/ML Engineer · 2× founder · ex-SpaceX · Caltech
APPLIED AI SYSTEMS
10+
years in production
Caltech
computer science
scroll · interact · ⌘K
00About

About me.

Sadaf Amouzegar

I'm an AI/ML engineer, builder, and systems thinker focused on turning complex ideas into practical, scalable products.

Over the last decade I've worked across startups, aerospace, entertainment, and life sciences, building everything from production ML infrastructure and LLM-powered systems to tools for budgeting, analytics, and large-scale data intelligence. My work tends to live at the intersection of engineering, research, and product design: taking messy real-world problems and turning them into systems people can actually use.

But outside of the technical side, I'm someone deeply motivated by curiosity.

I've always been drawn to understanding how systems work, whether that's machine learning models, human behavior, language, art, music, storytelling, or the ways people make decisions under uncertainty. I love structure and organization, but I'm equally driven by creativity and emotional depth. A lot of my best work comes from balancing both sides.

I care deeply about building technology responsibly. Not just systems that are powerful, but systems that are thoughtful, reliable, and grounded in reality. I'm especially interested in how AI can augment human creativity and decision-making rather than replace it.

When I'm not building, I'm usually learning something new: studying languages, reading about machine learning and new breakthroughs in AI, writing poetry, making music, designing side projects, or disappearing into a topic until I understand it completely.

At the core of everything I do is the same instinct: to build things that are meaningful, elegant, and genuinely useful.

01Career · Level Select

Where I've shipped.

01 / 07Life Sciences · Regulatory AI
active
2025 — Present · Remote
Basil Systems
Principal AI/ML Engineer

Building production AI systems for medical device safety intelligence.

  • Designed LLM workflows for analyzing FDA recall and adverse-event data
  • Built evaluation and regression-testing pipelines against historical recall outcomes
  • Developed backend services for retrieval, data processing, and model-assisted analysis
02 / 07Applied AI
complete
2025 · Remote
Dewey Labs
Staff AI Engineer

ML pipelines for LLM document analysis and classification, powering decision-support systems with reliability and interpretability built in.

  • Risk-scoring workflows focused on safe, explainable model behavior
  • Backend + product components across AI features (Python · Ruby on Rails)
  • Evaluation frameworks to improve trust in AI outputs
03 / 07Founder · 0 → Production
complete
2022 — 2025 · Culver City, CA
RivetAI
CEO & Head of AI/ML

Built an AI platform for film and TV pre-production, focused on budgeting, scheduling, and script analysis.

  • Led product and technical direction from early prototype to production
  • Built ML pipelines for script breakdowns, extraction, and production planning
  • Deployed model-serving infrastructure on GCP supporting production users
04 / 07Founder · Brand Analytics
complete
2020 — 2023 · Remote
Storyroom AI
Co-Founder & Chief Data Scientist

Built NLP systems for brand analytics, market intelligence, and audience sentiment.

  • Designed pipelines for trend detection and competitive analysis
  • Built NLP models for extracting signals from public text data
  • Led early product, data, and engineering direction
05 / 07Applied NLP · Generative AI
complete
2017 — 2020 · Culver City, CA
RivetAI
Principal Data Scientist

Built ML systems automating production workflows. NLP pipelines for script breakdowns and content generation; optimization for production planning.

  • NLP engines for automated script analysis and story generation
  • Predictive frameworks for production success metrics
  • Generative-AI research foundational to a granted patent
06 / 07Aerospace
complete
2013 — 2016 · Hawthorne, CA
SpaceX
Data Scientist · Software Engineer · Test Software Intern

Three-year ascent through avionics test software, internal tooling, and ML. Anomaly detection on engine and Dragon-vehicle test data.

  • Anomaly detection across engine and stage test data; EMI testing on Dragon
  • Models to prioritize and rank production issues by severity
  • StarTrekker camera focus algorithm; calibration + video acquisition tooling
07 / 07Origin · Research
complete
2011 · Pasadena, CA
Caltech
Aerospace Corp. SURF Fellow

Undergraduate research in Babak Hassibi's lab on Hyperbolic vs. Euclidean Embedding of Real-World Networks. MDS against Crovella's online greedy embedding into the Poincaré disk.

  • Implemented MDS + online greedy hyperbolic embedding in MATLAB
  • Evaluated on preferential attachment, small-world, and Erdős–Rényi graphs
  • Aerospace Corp. SURF Fellowship · co-mentored with Elizabeth Bodine-Baron
Read paper
02Stack · Stat Sheet

The toolkit, in production.

tier · 01
LLMs & Agent Systems
.001
  • LangGraph
  • RAG pipelines
  • Multi-agent workflows
  • Agentic failure mode management
  • Fine-tuning (LoRA, PEFT, RLHF)
  • LangChain · LangSmith
  • MCP
  • Hugging Face
tier · 02
ML & Modeling
.002
  • PyTorch
  • Evaluation & experimentation
  • Deep learning
  • NLP
  • Model optimization
  • MLflow
  • TensorFlow
  • scikit-learn · XGBoost
tier · 03
Infrastructure & Data
.003
  • Python
  • GCP (Cloud Run, BigQuery, Pub/Sub)
  • Vector databases (Pinecone, Qdrant, Weaviate)
  • FastAPI
  • CI/CD
  • AWS
  • Graph databases (Spanner, Neo4j)
  • Docker
03Press

Press & selected work.

09 entries
04Connect

Find me where the work lives.

© 2026 Sadaf Amouzegar
samouzegar.com