RESEARCH DIRECTIONS · REAL PROBLEMS, WORKING PROTOTYPES
Understand the problem.
Build a useful answer.
I study real-world workflows, frame the problem clearly and build practical AI and software prototypes to test what helps. Each direction explains the question, the approach and what I learned.
Independent portfolio and research projects—not products for sale.
01 / DOCUMENT INTELLIGENCE & RAG
Can an AI assistant make business knowledge easier to use?
Important answers are spread across documents, messages and pages, so people spend time searching before they can act.
BizBuddy explores a retrieval-based assistant that answers in plain language, cites the supporting source and shows how booking or hand-off could fit around human review.
What this prototype explores
- Retrieve relevant passages instead of guessing from memory.
- Show source links so an answer can be checked.
- Test how the assistant handles missing or unsupported information.
- Explore a booking flow without presenting it as a live product.
BizBuddy
AI business assistant Ready
02 / BUSINESS ANALYTICS & EXPLAINABILITY
Can a clearer picture lead to a better next move?
A busy operation can still hide what is profitable, what is overdue and where capacity is being used.
BizPulse explores an explainable dashboard that brings sales, costs, invoices and workload into one view, then turns the numbers into a question worth investigating.
What this prototype explores
- Compare sales with recorded direct costs instead of relying on one number.
- Surface unpaid invoices and workload signals in one place.
- Make the assumptions behind each metric visible.
- Use the dashboard to support a decision, not replace judgment.
OPEN RESEARCH QUESTION
What changes when a repeated workflow becomes a system?
Some problems do not fit a named product. This space holds the smaller experiments: mapping a workflow, connecting the data and testing a focused tool that people can actually use.
See the working prototypesHOW I APPROACH A PROBLEM
Start with the question, not the tool.
I start with the people, the process and the constraints. Then I choose the smallest experiment that can teach me something useful.
View Rezwan’s résuméFrame the real problem.
Understand what happens today, who is involved and where information gets lost.
Design the smallest useful experiment.
Choose what to build, what to measure and where human review belongs.
Build, test and explain.
Try the prototype on realistic examples, document the trade-offs and share what the results mean.
BEFORE WE TALK
A few things
you might be wondering.
Are these products for sale?
No. These are independent portfolio and research projects built to demonstrate how I investigate problems and engineer working systems.
What makes a project worth exploring?
A useful project starts with a real question, clear constraints and a way to inspect the result. I focus on problems where a prototype can teach something concrete.
Does every project need AI?
No. AI is useful when language, documents or uncertainty are central to the problem. A simpler program, database or workflow may be the better engineering choice.
How do you evaluate the prototypes?
I define the expected behavior, test realistic examples and document where the system succeeds, fails or needs human review.
What information do the demos use?
The public demos use curated or illustrative information. They do not represent client data, client results or private employment systems.
Who is behind the work?
You’ll find the engineering background, employment experience, research interests and project notes on the About Rezwan page.
FOLLOW THE WORK
See the problem.
Open the prototype.
Explore the demos, read the project notes and ask the portfolio assistant about the engineering behind them.