# Zaara Yakub

> Founding Engineer · New York

- Email: zaarayakub@gmail.com
- GitHub: https://github.com/zaaray
- LinkedIn: https://www.linkedin.com/in/zaara-yakub-879897217/

## About

I love building things that give people back the time to do what only they can do and what they truly love doing.

That usually means automating tedious grunt work, replacing legacy tools nobody enjoys using, giving people the tools to start building, or turning a complicated process into something intuitive.

I spent years learning to code, then watched AI become alarmingly good at it. I see my foundation in computer science as leverage. It helps me build faster, learn faster, and pursue ideas that once would have been too ambitious to take on alone.

I like working across data, backend systems, product, UX, and GTM. I want to understand the whole problem, and I’m most excited by challenges without obvious solutions, especially when insights from data and thoughtful experiments can reveal a path forward.

## Experience

### Steadphast — Founding Engineer

*Apr 2026 — Present, New York*

The interaction layer for the inbound agent channel.

- Building the visibility layer for AI agent traffic — how many agents are hitting your product, who sent them, and what they are trying to do.
- Leading design-partner outreach across travel, financial services, and commerce.

### Chromium Quant Research — Co-Founder & Founding Engineer

*Nov 2025 — Present, Remote*

Autonomous trading across prediction markets.

- Designed and shipped an autonomous real-time trading engine that sizes its own bets under market-conditional risk constraints and runs unattended 24/7 on AWS, returning 30% across 15,000+ trades on live Kalshi and Polymarket data.
- Built high-frequency pipelines turning raw JSON streams into structured price, volume, and volatility datasets at 100ms ingestion latency for trading strategies.

### AI PREMie — Software Engineer & Data Scientist, Intern

*May 2023 — Jul 2023, Dublin, Ireland*

Microsoft- and SAS-backed health tech reducing preeclampsia-related maternal mortality.

- Built a Python application to clean, format, and analyze 100+ raw mass-spectrometry files weekly — cutting researcher analysis time by 55%.
- Implemented statistical analysis identifying key peptide biomarkers between diseased and healthy patient groups, feeding directly into diagnostic testing.
- Owned the full development cycle from discovery to deployment, on-site with researchers and engineers in Dublin.

### Callsign — Data Scientist, Intern

*Apr 2022 — Oct 2022, New York*

JPMC- and Accel-backed cybersecurity SaaS at $50M+ ARR — HSBC, Lloyds, Barclays.

- Analyzed fraud detection model performance across 1M+ user interactions and transactions using Python, SQL, and AWS.
- Built interactive Pandas and Matplotlib dashboards surfacing fraud trends and customer behavior for business stakeholders.
- Partnered with Customer Success on retention analysis that shaped enterprise engagement strategy.

### Cold Spring Harbor Laboratory — AI / Machine Learning Research Intern

*2020 — 2021, New York*

The top-ranked molecular biology and genetics research institution in the world.

- Designed, implemented, and benchmarked 5 neural network architectures on genomic datasets, raising interpretability score 30% via hybrid convolutional-recurrent designs.
- Identified disease-linked genome variants, cited in an internal whitepaper and poster presentations.
- Selected for the James Watson Partners for the Future program — 9% acceptance rate.

### Lehigh University — B.S. Computer Science & Engineering

*Aug 2021 — May 2025, Bethlehem, PA*

- Coursework: Software Engineering, Data Science, Discrete Structures & Algorithms, Computer Architecture, Web Systems Development, Blockchain, Parallel Computing.
- Marketing Chair of MESU and BALU; performer with the Off the Record a cappella group.

## Projects

### Undupe (2025)

*Finding the same product hiding behind twelve different listings*

Cross-platform duplicate detection for e-commerce. Scrapes listings from Temu and AliExpress, then decides whether two products are actually the same item — even when the titles are rewritten and the photos are re-shot at a different angle.

- Semantic title matching with Sentence Transformers, so "wireless earbuds pro" and "TWS bluetooth headset" collapse into one item
- Perceptual image hashing catches visually identical products across re-cropped and re-lit photos
- FastAPI service with SQLite persistence and an automatic scraping + similarity pipeline

**Stack:** FastAPI, Python, React, Sentence Transformers, ImageHash, SQLite

### Water Health VR (2024)

*A VR field trip that ended up in Esri’s official documentation*

A virtual reality game teaching water health, built with the Jacobsburg Environmental Center. Players explore a real watershed rendered from live geographic data instead of a hand-modeled set.

- First known integration of offline ArcGIS map data into the Unity engine — the documentation request filed along the way was adopted into Esri’s official docs
- MVP playtested with 40+ users and iterated on their feedback
- Partnered directly with the Jacobsburg Environmental Center on curriculum accuracy

**Stack:** Unity, C#, ArcGIS SDK, VR

- [Read the publication](https://publications.immersivelrn.org/practitioner/article/view/422)

### Regulatory Genomics (2021)

*Teaching neural networks to read the non-coding genome*

Research at Cold Spring Harbor Laboratory on predicting regulatory activity from raw DNA sequence, and on making those predictions legible to the biologists who have to act on them.

- Designed, implemented, and benchmarked 5 neural architectures on genomic datasets
- Hybrid convolutional-recurrent design raised interpretability score by 30% over the CNN baseline
- Surfaced disease-linked genome variants, presented at the RECOMB/ISCB Conference on Regulatory & Systems Genomics

**Stack:** Python, TensorFlow, CNN + RNN, Genomics

- [Watch the poster talk](https://www.youtube.com/watch?v=enJ5Ahw0T1U)

### Pegasus (2025)

*Stroke rehab that talks to an Arduino pegboard*

A mobile rehabilitation app built for Assistive Technologies at Lehigh. Stroke patients complete pattern-matching exercises on a physical 6-peg board while the app tracks motor-skill recovery over time.

- React Native app pairs with a physical Arduino pegboard over serial, turning hardware input into session data
- Four difficulty levels with animated pattern previews so patients see the target before attempting it
- Persists level, pattern, and completion time per session for clinicians to review

**Stack:** React Native, Expo, Node.js, Express, MongoDB, Arduino

### The Buzz (2023)

*A social platform, shipped by a five-person team on schedule*

Full-stack social media platform spanning web and mobile, built as a scrum team and deployed to production.

- Integrated Google OAuth 2.0 and implemented the RESTful backend services
- Managed CI/CD through Bitbucket and ran the team on scrum, delivering the MVP on schedule

**Stack:** Java, React, Node.js, Flutter, ElephantSQL, Heroku

### Cookbook (2025)

*Paste a recipe URL. Get the recipe. Nothing else.*

A minimalist recipe aggregator that strips the life story, the ads, and the pop-ups off any recipe page — then lets you save what you keep coming back to.

- Parses structured JSON-LD metadata to pull title, ingredients, instructions, and yield cleanly instead of guessing at HTML
- Google OAuth sign-in with Pinterest-style collections and automatic deduplication

**Stack:** React, Node.js, Express, MongoDB, cheerio

## Skills

- **Languages:** Python, Java, TypeScript, JavaScript, C / C++, C#, Rust, R, SQL
- **Frameworks:** React, React Native, Node.js, FastAPI, Express, Flutter, Unity
- **Data & ML:** TensorFlow, PyTorch, Pandas, Matplotlib, Statistical modeling, Time series
- **AI & agents:** Agent orchestration, Tool use & function calling, Context engineering, Evals, Cursor, Claude Design, Factory, Orca
- **Infra & Tools:** AWS (EC2, S3), Docker, Dagger, Blacksmith, CI/CD, MongoDB, ElephantSQL, Heroku, Fly.io, Git, Bitbucket, Jira
- **Craft:** Scrum / Agile, UX design, Figma, Procreate, Canva, Proto.io, ArcGIS, REST API design
- **GTM:** Clay, HubSpot

## Interests

- **Singing** — All four years of college I sang with Off the Record, a co-ed a cappella group. We advanced to semifinals at ICCAs, where I won Outstanding Choreography for the entire set. I miss the rehearsals as much as the performances. [https://www.youtube.com/watch?v=lwiaRJFI-ZE]
- **Baking** — The only way to satiate my appetite for delectable goods is by baking them myself. I love sharing with anyone who is around, because a recipe’s yield is way too much for just me. The cake in the photo I made for my niece’s 4th birthday.
- **Drawing** — I used Copic markers in this one, but I also like using pencil, acrylic, graphite, oil, and watercolour.
- **Board games** — Wingspan has been my recent obsession. If you have recommendations for engine-builders with strategy, cute designs and small moving pieces, please share!

## Agent endpoints

- `https://zaarayakub.com/llms.txt` — Short profile plus a map of everything else, per llmstxt.org
- `https://zaarayakub.com/agent.md` — This entire site as one markdown document
- `https://zaarayakub.com/agent.json` — The same content as structured JSON
- `https://zaarayakub.com/agent-view` — Human-readable rendering of the markdown