System Architect · Software Engineer · Data Scientist
I’ve spent close to a decade leading engineering teams across travel, ride-hailing, and fintech — turning legacy platforms into reliable, high-impact systems. These days I lead the pricing team at Jabama, in Tehran, and spend most of my attention on the teams behind those systems.
I’m a technical leader with close to a decade of experience leading engineering teams and rebuilding large-scale systems — across travel, ride-hailing, and fintech. My work centers on building reliable platforms, improving delivery efficiency, and keeping technology aligned with product goals.
I care about clarity, solid engineering practices, and helping teams deliver with purpose and autonomy. I solve problems across software engineering, data, and machine learning, design the solutions, and lead the teams that build them.
Built the pricing team and architected the platform behind it. Designed and scaled the smart-pricing system — adopted by more than 10K hosts, generating roughly 60% of platform bookings, and lifting overall sales by over 10%. Developed the demand-prediction and booking-probability models that drive it.
Delivered the finance platform, automating contract payment calculations that had previously been manual. Designed and built the company’s experimentation platform, giving product and data teams a unified foundation for controlled experiments.
Led the pricing squad and turned the pricing service into a platform serving every product through a unified domain. Rewrote the price-calculator microservice with domain-driven design — 25% less resource use, ~2× faster feature delivery.
Built and led a team of five, architected an extensible suite of data-intensive business-intelligence applications, and oversaw a distributed ETL system processing a billion banking-transaction records.
An open-source library that brings static type checking to Python dataframes. Annotate a dataframe with its schema, and pykrete validates every column you touch — typos caught as you type, references tracked through filters, joins, and aggregations, with go-to-definition into the schema. Closes a familiar gap for data teams: the silent dataframe bug that only surfaces in production.
Designed as a closed-loop control system rather than a pure ML problem. The controller uses a PID-style loop with a conditional integral term, distribution-aware signal shaping, and bounded action sizing — control theory drawn directly from my electrical-engineering and robotics background. Pricing is usually treated as a prediction problem; here it’s treated as a control problem.
Many of the operational jobs a data team writes are state machines underneath — sequences of conditions, actions, retries, and timers — and all of them carry the same machinery for state, logging, idempotency, and recovery from failure. The Conductor is a framework that handles that machinery, so developers focus only on the business logic of each new job. Built for Jabama’s pricing governance, designed and developed following TDD.
For my master’s thesis: a GAN-based method for augmenting credit-card fraud-detection training data — synthesising realistic transactions to address the chronic class-imbalance problem that limits these models in practice.
Thesis: GAN-based data augmentation for fraud-detection models.
Project: design and implementation of a Brain–Computer Interface.
Whether you’d like to talk about engineering or leadership, share an idea, or simply introduce yourself, I’d be glad to hear from you. I read every message that comes in.
[email protected]