Professional product cases come first. Selected AI, automation, data, and research work is separated below as supporting evidence of implementation depth and feasibility judgment.
Three professional cases organized around the problem, evidence, trade-offs, delivery constraints, and outcomes that shaped the product decision.
Flagship B2B product case spanning an early-stage MMP from product discovery and specification through pilot-led validation, analytical scaling, ecosystem integrations, and initial commercial adoption.
~10 pilot/onboarding apps · 1st paying client confirmed · 1.2s → 0.25s recorded API example
Product and go-to-market case for moving a relationship-wellness product from prolonged pre-launch development into public release, behavioral analytics, monetization, and early market validation.
1,450 unique registered · 58 unique paying · IRR 22.1m recorded revenue snapshot
Enterprise product-judgment case centered on MVP scope, controlled field validation, and evidence-based technical acceptance of confidential vendor-delivered systems.
2 real ambulances in controlled EMS field test · 4 confidential systems under technical acceptance
Selected hands-on cases demonstrate product framing, technical feasibility, delivery, and operational judgment while remaining supporting evidence beneath the professional product portfolio.
AI-powered Telegram course advisor that interviews learners, assesses knowledge and goals, recommends a personalized module sequence, preserves context, and escalates unresolved cases to human support.
~500 starts · ~300 complete interviews · 3rd place · Bozhan educational product competition
End-to-end Telegram automation that extracts active opportunities, calculates Jalali deadlines, publishes a consolidated daily digest, and sends urgent reminders without routine manual content operations.
~500 members · ~2 months automated operation
Automated market-data and analytics pipeline for Iranian equities, combining daily snapshots, validated 1–10 year historical analysis, provider fallbacks, and read-only dashboard delivery.
Projects across AI, data, automation, software, and research — including academic NLP work. A focused preview is shown first; use the filters or “Show all” to browse the full archive.
Python batch messaging workflow for contact processing, validation, API integration, pacing, retry/failure handling, and structured execution reporting through an Android device/SIM gateway.
Python · API integration · pacing · failure handling · reporting
Telegram reservation and operations product used by a travel agency for passenger registration, tour selection, admin-controlled capacity approval, offline payment verification, and targeted updates.
~10 months real use · ~3 tours/week
Personal Python data product that combines Iranian market sources, price/fundamental inputs, a custom relative-performance formula, and Excel screening reports.
Practical screening workflow · 5 users
Personal Python monitor that checked one filtered Divar property URL every five minutes and sent desktop notifications with direct links when new listings appeared.
Two-person automated trading system that collected Binance/CoinEx market data, detected cross-exchange and triangular opportunities, validated price/volume constraints, and executed order sequences.
TensorFlow/OpenCV computer-vision prototype for compliance monitoring in a controlled organizational context, with human review.
TensorFlow · OpenCV · human review
Three-person Persian NLP research prototype coupling a Persian GPT-2 generator with a multi-task evaluator for poetic validity, structure, and target-poet style alignment.
Persian/English semantic medicine retrieval prototype combining Iranian FDA data with WHO ATC classifications, fuzzy matching, bilingual embeddings, and cross-lingual vector alignment.
65,469 enriched medicine records
Explainable Persian grammar-correction tool for He-Kasra errors using normalization, POS/lexical signals, embeddings, and rule voting/veto.
Transformer/BERT-based academic prototype that returns separate seven-level sentiment labels for eight aspects of Persian movie reviews.
Scrapy-based Persian text-mining pipeline that collected structured comments and metadata from selected Ganjoor poetry pages for exploratory NLP analysis.
Sharif University M.Sc. research combining deep feature extraction and similarity descriptors with CatBoost/gradient-boosting regression for drug–target binding-affinity prediction.
~10% average improvement reported in thesis
Individual Flutter/Django academic prototype for connecting drivers and passengers on planned intercity trips with shared travel cost.
Four-person hackathon concept for coordinating property-visit availability directly around a classifieds listing journey.