Hi, I'm Eshagh 👋

I build systems where every decision has an owner and a trail.
Who is allowed to decide this — and can you prove afterwards what was decided and why. Sometimes the decider is a dean. Sometimes it is a language model.

Aviation parts trading · University governance · Recruitment · Special-education care · Language & cognitive-skills education

Es

About

I lead product and engineering — architecture, the AI layer, and delivery. Recent builds include a multi-tenant aviation parts trading platform with LLM email-to-quote behind a per-tenant confidence gate; an AI-native applicant tracking system with guardrailed natural-language querying; a university course governance platform with per-field approval authority; and a care platform answering from families' own records with citations.

Skills

AI / LLM
LLM pipelines
RAG
pgvector
LangGraph
Vercel AI SDK
Anthropic Claude
Google Gemini
OpenAI
Multi-provider failover
Prompt versioning
Cost routing
Backend
Next.js 15
TypeScript
Node.js
PostgreSQL
Prisma
Drizzle
BullMQ
Microsoft Graph
Python
Platform
Multi-tenant architecture
RBAC & permission systems
Approval workflows
LiveKit / WebRTC
Real-time collaborative state
Docker
CI/CD
Chrome extensions
Payments
HLS video
Practice
AI-orchestrated development
Product engineering
KVKK / GDPR-aware data design
Selected Work

Systems I built end to end

Three of 8 production platforms — an inbox turned into a quoting pipeline, a university's course approval hierarchy, and an AI assistant that answers from a family's own records. Each has a full case study: what it does, how it is built, and what is deliberately not claimed.

Fenavia — AI-assisted trading platform for aviation parts
Aviation trade · B2B SaaSIn production · multi-tenant

Fenavia

AI-assisted trading platform for aviation parts

A multi-tenant B2B platform that runs an aviation parts trader's full commercial cycle: inbound RFQ, supplier sourcing, quote, sales order, invoice and airworthiness paperwork. Customer requests arrive as email; a queue-backed Microsoft Graph pipeline reads them with an LLM, matches inventory, and drafts a quote only when it clears a per-tenant confidence gate. Everything else routes to a human with the reason it was held.

Next.js
TypeScript
PostgreSQL / Prisma
Microsoft Graph
LLM extraction
Multi-tenant SaaS
Read case study
Course Governance Platform — Semester course approval for a university, from department head to Student Affairs
Higher education · internal toolsDelivered

Course Governance Platform

Semester course approval for a university, from department head to Student Affairs

Semester course planning used to live in spreadsheets and email, where authority differed by role and nobody could see what was waiting on whom. I built a role-based platform that carries every course from a department head's draft through dean and rector approval to Student Affairs — with permissions enforced per role, per faculty scope and per individual form field, multi-branch instructor assignment against live teaching load, and a full audit trail.

Next.js 15
React 19
TypeScript
PostgreSQL / Prisma
Approval workflows
RBAC
Read case study
Needa — AI care assistant with long-term memory
Healthcare · special educationSolo build

Needa

AI care assistant with long-term memory

A Turkish-language platform for families raising autistic children. Parents log structured daily observations and upload medical reports; an AI assistant answers caregiving questions using that history and cites the exact record behind every recommendation. Therapists and teachers join the same child's profile under a granular permission model. Built solo: data model, retrieval architecture, prompt system, background processing and cost controls.

Next.js 15
PostgreSQL + pgvector
Drizzle ORM
Vercel AI SDK
LangGraph
RAG / long-term memory
Read case study
Side Projects

I like building things

I have worked on 6+ innovative projects, exploring new technologies and creating impactful solutions. These projects showcase my passion for building meaningful applications that solve real-world problems.

  • Teknofsm

    Istanbul, Turkey

    STEM project development training and mentorship platform for high school students. Providing guidance and resources for innovative technology projects.
  • CytoDeep

    Istanbul, Turkey

    Deep Learning Based Autonomous Colony Counter. AI applications in microbiology, presented at 1st National Health AI Congress. Technologies: Python, OpenCV, PyQt5, NoSQL
  • Paksumatik

    Istanbul, Turkey

    Water Quality Analysis System analyzing water microbiological and chemical values with AI. System uses embedded systems, Arduino, image processing, sensors, and IoT technologies.
  • Autismo

    Istanbul, Turkey

    Robot-Assisted Autism Therapy. Face recognition and AI-supported education platform for children with autism.
  • E

    Epicontrol

    Istanbul, Turkey

    Bracelet project detecting epilepsy seizures - Regional second place in TÜBİTAK University Students Research Project competition.
  • E

    ETTS

    Istanbul, Turkey

    Voice-controlled mobility solution for disabled individuals. Technologies: Arduino, MIT App Inventor, IoT
Contact

Get in Touch

Want to chat? Just send me a message on WhatsApp and I'll respond whenever I can. I will ignore all soliciting.