Alhassan Alfarran

Full-Stack & Python Developer — Web & AI Systems

Open to offersMoscow · on-site / hybrid / remote

Web platforms & AI search, built end to end.

Python & full-stack developer in Moscow, building commercially since 2024: React + TypeScript up front, FastAPI or Node.js behind it. In 2026 I built search over an engineering-document archive (0.18 s p95, accepted by the client); in 2025 I was the lead developer of OKKP at Avenue Group.

Currently: open to Python and full-stack roles

See my workDownload CV

lower avg. response time on a critical endpoint: ~60%
internal system for 9 vehicle depots, as lead developer: OKKP
p95 search response over an engineering-document archive: 0.18 s
languages · EN C1 · AR · RU: 3

Currently: open to Python and full-stack roles

Experience and education

Elektroservis
·
Avenue Group
·
Junzi Tech Solutions
·
NUST MISIS
·
Ural Federal University

Alhassan

About

Arabic is my first language and English (C1) my professional one. I studied in Russia — a Bachelor's at Ural Federal University in Yekaterinburg, then a Master's at NUST MISIS in Moscow. I like owning a feature end to end: the data model, the API contract, the interface and the Excel export a non-technical colleague can run on their own.

The short story.

Where it started
In 2024, during my Master's, I started working commercially: freelance web apps for clients, then full-stack work at Junzi Tech Solutions.
Building with real teams
At Avenue Group I was the lead developer of OKKP, an internal quality-control system for 9 vehicle depots: I designed the architecture and API and wrote most of the code, in a team of three.
Latest project
In 2026 I built semantic and full-text search over an engineering-document archive for Elektroservis — FastAPI, an inverted index, sentence-transformers, FAISS and OCR — with a Master's from NUST MISIS (2024–2026) behind me.

How I work

No process theatre — just the habits I rely on.

Code review
Small, focused diffs. I comment on intent and trade-offs, not style, and block on correctness and clarity — not taste.
Definition of done
Tested behaviour, visible in production, documented for the next person, no known sharp edges left behind.
When things break
Contain the problem first, then write down honestly what happened. The fix is the easy part — changing the system so it doesn't repeat is the work.
Documentation
A short README per service: what it does, how to run it, where it can break. Updated with the change, not after.

Answers are quoted from my CV — nothing is generated.

Try

Experience

Commercial development since 2024: three companies and freelance work — from an internal system for 9 vehicle depots to search over an engineering-document archive.

Apr 2026 — Oct 2026
Elektroservis
Python Developer
Remote · civil-law contracts
Built a semantic and full-text search service over an engineering (PTO) documentation archive; accepted by the client.
Built a semantic and full-text search service over an engineering (PTO) documentation archive; accepted by the client.
- An engineer enters a cable mark, document code, equipment designation or a phrase and gets the document, the PDF page (Excel sheet and rows, Word section), a text fragment and the network path.
- Hybrid search with no LLM generation: a custom inverted index plus FAISS semantic search on sentence-transformers embeddings (all-MiniLM-L6-v2), behind FastAPI with a plain HTML/JS interface.
- Indexed PDF, DOCX, XLSX and RTF files — an archive of ~4.1k files, ~1.5k documents with text — with OCR of selected scanned pages (Tesseract + Poppler).
- Speed: code and mark queries from 9–10 s to 1.2–1.7 s, phrase queries from 55–70 s to 2–4 s; 0.18 s p95 after moving to the inverted index.
- Quality: 739 automated tests (pytest) and a 104-query control set with top-1/top-5 metrics. Runs as a systemd service on Ubuntu Server.
- Built with AI agents (Codex, Cursor, Claude Code): I gathered the requirements, wrote the agent briefs, reviewed the changes and verified them on the production index.
Stack: Python · FastAPI · sentence-transformers · FAISS · Tesseract OCR · pytest · systemd
May 2025 — Dec 2025
Avenue Group
Full-Stack Developer (lead developer on the project)
Moscow · team of 3
Lead developer of OKKP, an internal quality-control web system for a network of 9 vehicle depots.
Lead developer of OKKP, an internal quality-control web system for a network of 9 vehicle depots.
- Designed the architecture and API and wrote most of the code (~177 of 183 commits in 2025): React 18 + TypeScript (Vite), Node.js / Express 5, MongoDB.
- Violations and fines module with server-side filtering and pagination; modules for employees, driver call-backs, a Cloudinary document archive and in-app notifications.
- Per-depot KPI tiles and in-browser Excel export (SheetJS) of violations, employees and the depot summary; JWT sign-in.
- Deployment with Docker Compose on a Linux server; CI on GitHub Actions.
Stack: React · TypeScript · Vite · Node.js · Express · MongoDB · Cloudinary · Docker Compose · GitHub Actions
Oct 2024 — Jul 2025
Junzi Tech Solutions
Full-Stack Web Developer
Full-stack features for web applications with Node.js, Express, MongoDB and React.
Full-stack features for web applications with Node.js, Express, MongoDB and React.
- Cut the response time of a critical API endpoint by ~60% with caching and a redesigned database query.
Stack: Node.js · Express · MongoDB · React
2024 — Present
Currently
Freelance
Freelance Developer
Remote
Web applications for clients, alongside my studies and full-time roles.
Web applications for clients, alongside my studies and full-time roles.
- React / Next.js / TypeScript and Tailwind CSS web applications for clients.
Stack: React · Next.js · TypeScript · Tailwind CSS

Education

2024 — 2026
Master's — Computer and Information Sciences
NUST MISIS, Moscow
2020 — 2024
Bachelor's — Computer Science
Ural Federal University, Yekaterinburg

Certifications

Server-side Development with Node.js, Express and MongoDBCoursera
Introduction to SQLCoursera, University of Michigan
Understanding and Visualizing Data with PythonCoursera, University of Michigan
Machine Learning for AllCoursera, University of London
Programming with Google Go SpecializationCoursera

Selected work

Projects from my roles — private code, so you see the architecture — plus open-source prototypes built with AI agents, with live demos.

Work projects

Work project · Avenue Group

Lead developer on the project · 2025

OKKP Operations Platform

An internal quality-control web system for a network of 9 vehicle depots. I designed the architecture and API and wrote most of the code: violations and fines, employees, a document archive and Excel export.

Result: Deployed on the company's Linux server; staff filter violations and export reports to Excel on their own.

Tech stack: React · TypeScript · Vite · Node.js · Express · MongoDB · Cloudinary · SheetJS · Docker Compose · Tailwind CSS

Read case studyPrivate code · NDA

Work project · Elektroservis

Python developer · Apr–Oct 2026 · semantic + full-text search

Engineering Document Search

Semantic and full-text search over an engineering-documentation archive of ~4.1k PDF, DOCX, XLSX and RTF files: a custom inverted index plus FAISS on sentence-transformers, behind a FastAPI service. No LLM generation.

Result: Accepted by the client in September 2026: 0.18 s p95, 739 tests and a 104-query control set.

Tech stack: Python · FastAPI · FAISS · sentence-transformers · Tesseract OCR · pytest · systemd

Read case studyPrivate code · NDA

Open-source projects

Open-source platform, built with Claude Code · Gateway in front of LLM inference servers

Switchyard

A gateway in Go, standard library only, that sits in front of LLM inference servers: it never sends a replica more than it can run, refuses work it cannot serve in time with a retry time instead of a timeout, and retries a failed request on another replica only while the client has seen nothing - measured on a real GPU against a plain load balancer.

Built in: 78 of 212 requests on time at 1.0/s against 6 for a plain proxy, no failures against 150; no client errors with a replica killed under load; a live dashboard, 47 tests and CI.

Tech stack: Go · llama.cpp · Ollama · Prometheus · Docker · Caddy · Python · GitHub Actions

Read case study Live demo GitHub

Open-source platform, built with Claude Code · Evaluation platform for tool-using LLM agents

Proofbench

An evaluation platform for an LLM agent that moves money: deterministic graders read what the agent actually did, an LLM judge is used only for prose and is itself measured against a blind panel, recorded runs re-grade in seconds without a GPU, and a CI gate fails when grading quietly gets weaker.

Built in: 75.0% fully correct and 95.8% safe at temperature 0; judge kappa 0.93 against a blind panel; re-grading in 4 s instead of 18 minutes; a live dashboard, 171 tests and CI.

Tech stack: Python · FastAPI · Pydantic · Ollama · NumPy · pytest · Docker · GitHub Actions · Caddy

Read case study Live demo GitHub

Open-source platform, built with Claude Code · Payments platform built like a bank's

Ledgerline

A payments platform where money is never created, lost or moved twice — a double-entry ledger on an event log, card payments and payouts as sagas, an outbox into Kafka, in-line fraud scoring, ClickHouse analytics, a gRPC merchants' API, and OpenTelemetry tracing, all proven by breaking it on purpose.

Built in: Nothing lost or done twice over 11 chaos scenarios and 40,461 requests; fraud caught 88% vs 37% for rules; ~430 ledger transfers/s on one CPU; every alert proven to fire; a live demo and CI.

Tech stack: Kotlin · Spring Boot · PostgreSQL · Kafka · ClickHouse · gRPC · Python · LightGBM · Vue · OpenTelemetry · Grafana

Read case study Live demo GitHub

Open-source platform, built with Claude Code · Live digital twin of a vehicle fleet

Depotwin

A live 3D digital twin of 240 simulated vehicles in Moscow — a Go and NATS JetStream pipeline into TimescaleDB, failure models that warn of breakdowns days ahead, a local-LLM copilot over read-only queries, and an offline-first driver app.

Built in: Overheating caught 4.5 days ahead (rules: 3.0) with fewer false alarms; 17 of 18 held-out copilot questions right; 200 requests/s with no errors under load; a live demo and CI for every part.

Tech stack: Go · Python · TypeScript · NATS JetStream · TimescaleDB · FastAPI · React · three.js · Expo · LightGBM · Prometheus

Read case study Live demo GitHub

Open-source platform, built with Claude Code · AI recruiting with evidence

Clearhire

An AI recruiting platform that checks CVs against each requirement of a vacancy and shows the CV line behind every verdict — blind screening, a local LLM that must cite its sources, and an audit log anchored on a blockchain.

Built in: nDCG@10 0.972 on 312 synthetic CVs (embeddings alone: 0.603); identical scores in all 486 blind-mode swaps; a live demo, and CI covering Python, three front ends, the contract and a Kubernetes end-to-end run.

Tech stack: Python · FastAPI · PostgreSQL · pgvector · Angular · Next.js · NestJS · Solidity · Kubernetes

Read case study Live demo GitHub

Open-source prototype, built with AI agents · Hybrid document search (RAG)

Lumen Search

Hybrid document search with citations over PDF, Word, Excel and scanned pages, in English and Russian — BM25 plus multilingual embeddings in FAISS, OCR and a FastAPI service.

Built in: 83.9% of 62 labelled questions get a correct passage in the top five after reranking. Live demo in the browser; tests on three Python versions and a Docker smoke test in CI.

Tech stack: Python · FastAPI · FAISS · sentence-transformers · Tesseract OCR · Redis · Docker

Read case study Live demo GitHub

Open-source prototype, built with AI agents · Uptime & SLO monitoring

Pulseboard

Self-hosted HTTP uptime monitoring with automatic incidents, SLO error budgets, a public status page and Prometheus metrics — all in one container.

Built in: Tests cover incident recovery, SLO math, access control and SSRF protection. Live demo; one docker compose command.

Tech stack: Python · FastAPI · SQLite · Docker · Prometheus

Read case study Live demo GitHub

Open-source prototype, built with AI agents · Release operations & DORA metrics

DeployLedger

A self-hosted release-operations control plane: it ingests deployment events, computes DORA metrics and helps decide whether the next release is safe to ship.

Built in: Computes all five DORA metrics from deployment events, with a tamper-evident audit chain. Live demo; CI tests the code and scans the images.

Tech stack: Python · FastAPI · React · PostgreSQL · Docker · Terraform · Prometheus

Read case study Live demo GitHub

Open-source prototype, built with AI agents · Just-in-time access control

Gatehouse

A self-hosted access request and approval control plane: narrowly scoped, short-lived access, with every decision recorded in a verifiable audit chain.

Built in: Idempotency keys, optimistic locking and a hash-linked audit chain per workspace. Live demo and public OpenAPI.

Tech stack: Python · FastAPI · React · TypeScript · PostgreSQL · Docker · Terraform

Read case study Live demo GitHub

Open-source prototype, built with AI agents · Webhook debugging in Go

Webhook Workbench

A private, self-hosted workbench to capture, inspect, verify and safely replay webhooks — one dependency-free Go binary with the UI built in.

Built in: Tests cover redaction, signature checks, replay and SSRF defences. Checksummed release binaries; live sandbox.

Tech stack: Go · net/http · Docker · HMAC-SHA-256

Read case study Live demo GitHub

Full project archive

Performance deep dive

Junzi Tech Solutions

Cutting a critical endpoint's response time by ~60%.

Queries first, cache second: I refactored the MongoDB queries behind a critical Node.js / Express endpoint, then cached repeated reads, without changing its response contract.

Result

~60%

lower average response time on a critical API endpoint

Average response time, critical endpoint (relative scale)

01Problem
A critical endpoint in a client web application (Node.js / Express on MongoDB, React front end) was too slow, and the fix had to keep its response contract unchanged.
02Solution
I refactored the database queries behind the endpoint and added a caching layer for repeated reads. Validation, error handling and the response contract stayed consistent with the rest of the REST API.
1. Client
  React
  : React front end.
3. API
  Express
  : Node.js / Express endpoint with consistent validation and error handling.
5. Cache
  hit · returns early
  : Caching layer for repeated reads.
6. miss
7. Data
  MongoDB
  : Refactored MongoDB queries.
03How I approach performance work
- Measure first — the real cost is rarely where you look first.
- Fix the query, then cache: a cache on top of a slow query only hides the problem.
- Keep the response contract stable, so a performance fix ships like any other change.

Capabilities

What I work with, grouped by layer — from interfaces to delivery.

Frontend: React, Next.js, TypeScript, JavaScript, Tailwind CSS, Vite, Framer Motion
Backend: Python, FastAPI, Node.js, Express, REST API design, JWT authentication, File uploads (Multer)
Search & Documents: sentence-transformers, FAISS, Inverted index, Hybrid search, OCR (Tesseract, Poppler), PDF / Word / Excel parsing, LLM / RAG (prototypes, built with AI agents)
Data & Storage: MongoDB (Mongoose), Cloudinary, SQLite / PostgreSQL (university project), pandas & NumPy (university projects)
Delivery & Tools: Git & GitHub, Docker / Docker Compose (basic), GitHub Actions, Linux (Ubuntu, systemd), pytest, Vercel, Claude Code · Cursor · Codex

Services

For teams that need an internal platform, a reporting dashboard or search over their own documents — designed and built end to end, from the database to the UI. Contract or freelance, remote or in Moscow.

Full-stack web platforms
Admin panels and internal tools with authentication, server-side filtering, file uploads and clean REST APIs — like OKKP.
Example: OKKP
Document search
Search over PDF, Word, Excel and RTF archives: OCR for scans, an inverted index plus semantic search, FastAPI services.
Example: engineering document search
Dashboards & reporting
KPI tiles, filterable tables and Excel exports your team can use without asking a developer.
Example: OKKP reporting
API performance & backend
Profiling slow endpoints, caching and query refactoring, consistent validation and error handling.
Example: the ~60% case study
Multilingual & RTL interfaces
English, Russian and Arabic UIs with real right-to-left layouts, like this site.
Example: Smart Platform (EN / AR)

Discuss a project

3D web

Real-time 3D sites for brands, built performance-first. Every model, shader and particle is made in code — there are no model files to download.

Render on demand
A frame is drawn only when something changes, so an idle page costs almost nothing.
Adaptive quality
Resolution and effects scale to the device to keep motion smooth, phones included.
Reduced motion
With reduced motion turned on, scenes settle into a still view that stays fully usable.
Made in code
Geometry, shaders and particles are procedural: nothing to load but the code itself.

Stack

Next.js
React Three Fiber
Three.js
GLSL
GSAP

Live
Eco-city developer
Skybloom (Live)
Glass towers rise from the ground as you scroll, greenery spreads across them and the light shifts from day to dusk.
GitHub
Live
Specialty coffee house
Qahwa House (Live)
A 3D cup with a live pour, rising steam particles and coffee beans in orbit.
GitHub
Live
Beauty & nail studio
Bloom Studio (Live)
An iridescent polish bottle that the colour picker repaints in real time with a thin-film shader.
GitHub
Live
AI assistant SaaS
Nova AI (Live)
A GPU particle “mind” orb that reacts to the cursor and to what you type.
GitHub
Live
Luxury real estate
Horizon Tower (Live)
A procedural tower with a floor selector, a day/night toggle and a sun-angle slider.
GitHub
Live
Watch configurator
Aurum (Live)
A procedural watch with case, strap and dial pickers and an exploded view.
GitHub

GitHub

Prototypes, experiments and earlier work — hand-picked and grouped by area. Many of the recent ones are prototypes built with AI agents (Claude Code, Codex).

@kyan9400

Live
Stockroom Ledger (Live)
Transactional inventory operations with audited cycle counts, idempotent writes, and optimistic concurrency.
Python
GitHub
Live
MyGPT (Live)
Readable character-level transformer with reproducible training, safe checkpoints, and an interactive causal-attention explorer.
Python
GitHub
Live
LeaseQueue (Live)
Durable background jobs with atomic leases, retries, idempotency, and dead-letter recovery.
Python
GitHub

Full project archive

Contact

Hiring, or have a project? Let's talk.

Hiring for a full-stack or Python role — or planning a web app, an internal tool or search over your documents? Tell me what you need. Telegram is the fastest way to reach me.

Telegram@hassan_9400

Email mekyan775909@gmail.com

Download CV

English Русский العربية

Open to offersMoscow on-site/hybrid or remote · full-time or contract

I reply within 48 hours — faster on Telegram.

LinkedIn GitHub

Web platforms & AI search, built end to end.