● open to opportunities

Hi, I'm Maryem.

I'm a software engineer in Basel. I spent six years making backends behave, and lately I've been teaching large language models to do genuinely useful things. I'm so glad you're here. Scroll down and I'll show you around.

scroll to meet me ↓
// about_me

I build the backend. Then I taught it to talk.

  • ▸I spend my days in the backend. APIs, servers, the quiet stuff that just has to work.
  • ▸I care about clean code and solid tests. Clean code the next person is happy to open, and unit and functional tests that keep the system reliable.
  • ▸A few years ago I got really into large language models, so I did my Master's part time while working full time.
  • ▸My thesis was about getting LLMs to read privacy policies for you.
  • ▸At work I put AI into real products, like an MCP server that lets an AI call our APIs on its own.
maryem — whoami
$ cat about.json
{
"location": "Basel, Switzerland",
"role": "Software Engineer @ Peerdom",
"experience": "6+ years",
"currently": "fine-tuning LLMs",
"degree": "MSc CS, Uni Basel",
"work_permit": "B (Switzerland)",
"status": "open to good problems"
}
// where_I've_shipped

Six years, three teams.

Software Engineer · Peerdom

04/2023 — now · Bern, CH
  • ▸Built an HTTP MCP server that registers REST APIs as typed, LLM-callable tools — any AI client can run multi-step org tasks from natural language.
  • ▸Integrated OpenAI LLMs to automate resource creation, cutting delivery from minutes to seconds.
  • ▸Designed clean, modular backend architecture in TypeScript / Node.js; led system-design discussions and reviews.

Application Software Engineer · Motive

11/2019 — 02/2022 · Islamabad
  • ▸Core engineer on an ELD compliance platform for US & Canadian fleets — Hours-of-Service logging and PDF reporting in Ruby on Rails.
  • ▸Extended the platform from US FMCSA rules to Canadian HOS (Transport Canada), translating federal regulations into production logic.
  • ▸Wrote unit & functional tests for CI, and ran on-call support for live production issues.

Application Software Engineer · An10

06/2018 — 11/2019 · Islamabad
  • ▸Shipped RESTful APIs with Express.js / Total.js and Sequelize across MongoDB and MySQL.
  • ▸Used the Elasticsearch Document API for fast ad-hoc queries over large-scale IoT sensor data.
  • ▸Designed the JWT + ACL auth layer and broke components into Dockerized microservices.
// master's_thesis · Uni Basel

Can an LLM read the privacy policy so you don't have to?

Nobody reads privacy policies — 74% of people skip them entirely, and reading every one you agree to would cost ~244 hours a year. My thesis asked a sharper question: can large language models catch the moment your data gets handed to a third party, reliably enough to trust?

5
LLMs benchmarked head-to-head
13
personal-info types detected
2
datasets — OPP-115 & MAPP
+0.26
F1 lift from QLoRA fine-tuning
finding_01

Structured output + low temperature (0) was the single most reliable combo. Discipline beat cleverness.

finding_02

Prompt design gave bigger gains than fine-tuning for most categories. Still, QLoRA lifted Mixtral from 0.53 to 0.70 F1 on cookie-tracking.

finding_03

Models nailed concrete labels ("cookies & tracking") but stumbled on vague ones like "personal information."

models_in_the_ring:
DeepSeek-R1-Distill-Qwen-32B Mixtral-8x7B Llama-3.1-8B Grok 3 Gemini 2.5 Flash
↓ Read the thesis (PDF) ↓ See the slides Thesis code on GitHub ↗
// my_toolbox

Things I reach for.

AI / LLM

Fine-tuning LoRA / QLoRA Prompt engineering MCP LLM eval PyTorch Hugging Face LangChain OpenAI API Vertex AI

Languages

TypeScript JavaScript Python Ruby Java C / C++

Data & infra

PostgreSQL Elasticsearch MongoDB Redis Neo4j Docker GitLab CI/CD Datadog
// say_hi

Let's build
something good.

Hiring, collaborating, or just want to talk LLMs and clean backends? My inbox is open and I'd love to hear from you.

open to opportunities · Basel or remote
✉ maryem.fatima96@gmail.com in / LinkedIn ↗ GitHub ↗
download: CV ↓ Thesis PDF ↓ Thesis slides ↓
MARYEM FATIMA · SOFTWARE ENGINEER · LLMs · BASEL ·  MARYEM FATIMA · SOFTWARE ENGINEER · LLMs · BASEL · 
© 2026 Maryem Fatima · built with love, pink, and a little Python