sz.

A little about me

Hi, I’m Sam
Zakharia.

Curious by nature.
Practical by approach.

I build tools that help people make sense of information and get things done. My path has taken me through chemistry, running a business, and hands-on technology leadership.

Explore my work
Sam Zakharia smiling by the ocean
Sam, away from the keyboard.
Science → Business → TechnologyA few different perspectives. One curious mind.

01 / Selected work

Useful things,
built thoughtfully.

Some projects help a team work better. Others start with an idea I want to explore. Here’s a look at both.

01AI & researchCompleted

A conversation with
supplement science.

A RAG platform for exploring research on supplements and nutrients—from a topic across the entire library to a conversation with a single paper.

Inside the project

What it does: I built a research workspace that automatically brings new supplement and nutrient papers into a growing library. People can search the whole collection by topic, discuss a single paper, or group several papers into a shared source for a conversation.

The problem: Relevant research can be spread across many papers. Finding a topic is only the first step; people also need a way to explore the material in detail and consider related papers together.

The approach: Split research papers into appropriately sized chunks, convert those chunks into vector embeddings, and store them in a local Milvus database. Retrieve relevant material from the selected sources and use OpenAI models to analyze it and support the conversation. Automatic ingestion keeps new papers coming into the library.

The technical focus: Supporting different scopes of research: a broad search across the library, a close reading of one paper, or a discussion grounded in a selected group. Source selection and the context passed to the model are central to making those modes useful.

  • RAG
  • Milvus
  • Vector embeddings
  • OpenAI models
  • Research ingestion
  • Library-wide search
  • Grouped sources
02Internal softwareCompleted

Company documents.
Kept within the company.

An internal PDF workspace for editing, annotations, collaboration, and approvals, with documents stored inside the company network.

Inside the project

What it does: I built an internal document workspace with familiar Acrobat-like tools. Colleagues can edit and annotate PDFs, collaborate on changes, store documents within the company network, and move them through an approval workflow.

The problem: Document work includes both making changes and getting the right people to review them. The system brings those steps together while keeping the underlying files inside the company.

The approach: A Flutter frontend provides the workspace, with Syncfusion handling PDF functionality and Python powering the backend. The approval workflow connects with Microsoft 365 Outlook to notify the people involved when their attention is needed.

The boundary: Keep the documents entirely within the company network, without exposing them to the internet. Core document work operates internally; Outlook handles approval notifications while the documents themselves stay inside the company.

  • Flutter
  • Syncfusion PDF
  • Python
  • PDF collaboration
  • Approval workflows
  • Microsoft 365 / Outlook
  • Internal storage
03Advertising analyticsCompleted

Better keywords.
More considered spending.

An Amazon advertising platform for finding long-tail keywords and evaluating both advertising efficiency and financial performance.

Inside the project

What it does: I built an Amazon advertising platform that generates long-tail keyword opportunities and helps evaluate campaign spending. Users can consider ad placement and performance alongside costs, revenue, and contribution margin.

The problem: Competing for expensive keywords against stronger advertisers can make growth costly. Sales and clicks alone also leave out an important question: what does the business keep after its costs?

The approach: Use Amazon Search Query Performance (SQP) and advertising data to identify more specific search opportunities. Bring those opportunities together with financial measures so keyword and spending decisions reflect both advertising efficiency and business performance.

The technology: Amazon Selling Partner APIs (SP-API) and Python data analysis support the platform. Connecting search and campaign data with the economics behind a sale helps evaluate long-tail keyword opportunities alongside contribution margin.

  • Long-tail keywords
  • Amazon SP-API
  • Python data analysis
  • Amazon SQP
  • Advertising efficiency
  • Contribution margin
04ERP & intelligenceWIP

A clearer view of
the CPG business.

An ERP and intelligence platform in development, bringing operational planning and customer insight into one consumer packaged goods (CPG) software suite.

Inside the project

What it does: I’m building an ERP and intelligence suite for a consumer packaged goods (CPG) business. It is intended to help teams manage reorder points, understand customer trends, and use that information in day-to-day planning.

The problem: Inventory decisions and customer behavior are connected. Planning benefits from seeing when stock needs replenishing alongside how demand is changing, within a broader view of the business.

The approach: Bring enterprise resource planning and business intelligence into one software suite, with reorder points and customer trends as core areas of focus. My experience running a business helps shape the workflows around practical decisions.

The ongoing challenge: Turning a broad CPG intelligence goal into useful operational tools. This is still a work in progress: the focus is on connecting information to decisions people need to make, rather than presenting it as a finished system.

  • ERP
  • Reorder points
  • Customer trends
  • CPG intelligence
05Personal projectLive

A place for
imaginary worlds.

DunMax — Dungeoness Maximus — is my AI worldbuilding app for creating characters, organizing NPCs, and developing connected worlds.

Inside the project

What it does: DunMax helps people create characters, organize NPCs, and develop connected fantasy worlds. I rebuilt it from my earlier Fantasy World Builder app and brought it to the App Store and Google Play.

The problem: Worldbuilding involves keeping many pieces of lore connected. Building the app also introduced a different challenge: giving people a responsive experience when some AI requests take much longer than others.

The approach: Separate immediate API requests from time-intensive background work. Google Cloud Pub/Sub passes work to background processing, so a long-running task does not hold up the immediate API response. Redundancy measures, database backups, and data archiving address reliability and preservation of information.

What I learned: My first app did not use asynchronous operations, so slow work could hold everything else up. Rebuilding it taught me to think more carefully about APIs versus workers, and about what happens when something fails. Protecting an internet-facing service and its data became part of the development work, alongside the features people see.

  • Google Cloud Pub/Sub
  • APIs & workers
  • Redundancy
  • Internet security
  • Backups & archiving
App Store Google Play

02 / How I work

The tools are varied.
The goal is practical.

I work across the software itself, the infrastructure behind it, and the people who need to use it.

01AI & knowledge systemsModels · Retrieval · Experimentation

I got involved early in the generative AI wave and have followed its evolution through hands-on experimentation ever since. My interest spans language models, local inference, image and video generation, coding assistants, and agent tools. The tools keep changing; I enjoy understanding what each new generation makes possible and putting it to use.

I work across OpenAI, Anthropic, and Google models, choosing different models for different purposes. I keep up with new releases, changing capabilities, and the recurring “almost at AGI” conversation, while paying attention to what actually works in practice. Context, token usage, and the needs of the task all shape how I use a model.

Building retrieval systems from the underlying pieces—ingestion, chunking, embeddings, and vector search—gave me a foundation for understanding the platforms that now package those techniques together. My Milvus research project is one example of that experience. Knowing how the pieces evolved helps me evaluate new tools and understand what is happening underneath them.

That same curiosity has taken me through tools such as Midjourney, OpenClaw, and Claude Code, alongside managing Claude Enterprise accounts. They are examples along an ongoing learning path. I’m always exploring what comes next and where it can solve a useful problem.

02Software & cloudPython · Flutter · Google Cloud · Go

I use Python for backend services and data analysis, and Flutter for application interfaces. My projects include Syncfusion PDF tools, Amazon Selling Partner API integrations, and systems that connect AI models with useful business workflows.

On Google Cloud, I’m familiar with Artifact Registry, IAM permissions, service accounts, Cloud Run services and jobs, Cloud Scheduler, Secret Manager, and Pub/Sub, as well as Firebase Authentication and Hosting.

I use Go as an alternative to Python for simple jobs where fast container cold starts matter. Building DunMax also taught me to separate immediate API responses from background work, and to account for security, redundancy, database backups, and archiving.

03Business & scientific thinkingProblem solving · Operations · Learning

Problem solving comes first. Tell me what is getting in your way, and I’ll work with you to find a solution. I like understanding the process behind a problem before deciding which tool belongs in it.

My path has covered chemistry, running a business, and building software. Each step taught me something: testing assumptions through scientific method development, understanding costs and operations, and turning an idea into a system people can use.

I know that is a varied background, but I bring those lessons with me. I want work that challenges me, gives me something new to learn, and lets me make a practical difference.

04Enterprise technologyMicrosoft 365 · Entra · Exchange · Azure

I’ve managed the Microsoft 365 environment, including Entra, Exchange, and Azure. That experience sits alongside managing enterprise AI accounts and working with outside technology vendors.

It gives me a view of technology beyond an individual application: the services a company relies on, the people who need access, and the vendors supporting the work.

03 / The path here

A background
with a few
different chapters.

Each role has given me a different way to look at a problem.

Education

B.S. in Chemistry

American University of Beirut
Beirut, Lebanon

Graduated 2012 · Top 10% of my class

Joined 2024

Director of Intelligent Tech

CAS BioSciences

Built a RAG platform for supplement and nutrient research, an Amazon advertising and financial analytics platform, and an internal PDF workspace with approval workflows. Currently developing an ERP and CPG intelligence suite for reorder planning and customer trends.

Alongside development, I’ve managed Microsoft 365 and enterprise AI accounts, worked with outside technology vendors, and built smaller AI projects.

October 2016 — April 2024

Entrepreneur

Barbonzai · Garbanzai Bakeries LLC

Launched and operated the business, covering financial planning, hiring, payroll, vendor relationships, and standard operating procedures. Established a second location, including construction requirements and permits. Stepped back in April 2024 and gradually wound down the company.

From a restaurant to a different kind of building.

I wanted to start a business and made the age-old mistake of opening a restaurant. I love cooking, and I’m a very good cook. But loving the food and keeping a restaurant afloat are different things. COVID hit us hard, and we never properly recovered.

How that led me to code ↓

During that period, I got into the AI boom and started teaching myself to code. AI felt like having the world at my fingertips. I built my own learning syllabus, starting with Python and small, sometimes silly programs.

That grew into Fantasy World Builder, my first app for making D&D worlds. I coded it myself and made plenty of beginner mistakes along the way—including not using asynchronous operations, so slow work could hold everything else up. A lesson learned the hard way.

Later, I rebuilt it from the ground up as Dungeoness Maximus. That progression—from trying things to understanding why they worked—is a big part of how I learn.

March 2015 — May 2016

Chemist II

Best Formulations

Developed and validated a USP <561> pesticide assay using GC-MS within three months. Performed quality-control testing with GC, HPLC, FTIR, UV-Vis, and TLC.

September 2013 — March 2015

Chemist I

Captek Softgel International

Developed fatty-acid assay methods for gas chromatography. Worked with HPLC, GC, UV assays, titration, pH, and viscosity testing.

Sam Zakharia standing beside the ocean
There’s always something else to learn.

04 / Outside the job description

Still curious,
off the clock.

I like learning things. Sometimes it’s a useful new skill. Sometimes it’s a completely random piece of information that sends me down a rabbit hole. Gemini has been a great companion for that curiosity: helping me explore follow-up questions and keep digging into a subject that catches my attention.

Finance & markets

I’m interested in markets, the psychology behind financial decisions, and cryptocurrency. I enjoy exploring how people, expectations, and information shape what happens next.

Learning, in any direction

Building PCs, plumbing, or a completely random subject—I like figuring things out. With so much information within reach, I believe there’s always a way to start learning something new.

Games that put up a fight

Elden Ring and Dark Souls are my favorites. I enjoy learning the mechanics, trying again, and finally getting past the part that seemed impossible.

Cooking, with confidence

I love cooking, and my best dishes include ceviche and wara2 3enab (stuffed grape leaves). The restaurant chapter ended; the enjoyment of making good food didn’t.

05 / Say hello

Interested?
Let’s have a conversation.

If my experience sounds relevant to your team, I’d be happy to talk.

Your message goes straight to my inbox.

Received, thank you.

Thanks for getting in touch. I’ll read your message and get back to you.