Moatasim Bin Hisham Sayyid — Software Engineer (Backend & Applied AI)

I build production backend systems, ship AI people actually rely on, and I have a soft spot for maps.

Three years at Folio3 turning multi-gigabyte CAD drawings into interactive maps, wiring enterprise .NET services together, and running a RAG assistant 700+ colleagues use daily. Karachi-based, open to relocating.

status

open to roles, incl. relocation

now

Software Engineer, Folio3

primary

C# · Python · TypeScript

based

Karachi, PK · UTC+5

$ cat focus.txt

Backend & distributed systems

Production services in C#/.NET and Python. Event-driven architectures, Clean Architecture, DDD — designs that stay legible after the third feature request.

Applied AI & LLM tooling

RAG, Model Context Protocol servers, LLM evaluation. I care about AI being trustworthy, not just impressive — which mostly means measuring how it fails.

Geospatial

CAD/GIS pipelines, coordinate reference systems, rendering big spatial data. Centimetre-level accuracy is a surprisingly good teacher.

$ ls ./projects

4 featured

01
flagship

LandTeller

landteller.com ↗

A geospatial platform that turns multi-gigabyte AutoCAD (DWG) and shapefile drawings into interactive Mapbox tilesets. I lived in coordinate reference systems and centimetre-level accuracy, integrated GDAL and ODA to convert CAD geometry into universal GeoJSON, and ported parts of a C++ SDK into native C# just to get proper in-process debugging.

When the pipeline buckled at scale I re-architected it as a fully event-driven AWS system — a Step Functions state machine of 1,200+ lines of ASL, with S3 → EventBridge → SQS → Lambda. End-to-end processing latency dropped 77%.

C# / .NETAWS Step FunctionsLambda · SQS · S3GDAL / ODAMapbox

02
thesis

Recruit Ranks

repo ↗

An NLP system that scores how well a CV matches a job description, 0–100 with a written rationale. It fuses embedding similarity with a cross-provider LLM debate: ChatGPT and Gemini argue opposing cases and a third model judges. An early, slightly stubborn take on LLM-as-judge — built before that was a phrase people used.

PythonEmbeddingsLLM evaluation

03
product

Khelgah

A multi-tenant sports-facility booking platform — NestJS/Fastify backend, React Native app, Postgres and Prisma. Redis hold-locks keep two people from booking the same pitch at the same minute, which is the entire product in one sentence. Built by someone who has spent a lot of time trying to find a pitch in Karachi.

NestJSReact NativePostgres · PrismaRedisDocker

04
research

GRC Automaton

repo ↗

A fork of an autonomous-agent framework where agents hold wallets and must earn their own compute. I tested whether an LLM agent could trade unsupervised on a testnet, then built the guardrails and eval loops to study how it fails. The failures were the interesting part.

PythonAgentsGuardrails & evals

$ cat stack.toml

[backend]

C# · ASP.NET Core
Python · FastAPI
Node.js · NestJS
Django REST
Clean Architecture · DDD
RabbitMQ · SQS

[applied ai]

RAG pipelines
Model Context Protocol
FastMCP · LangChain
Embeddings
LLM evaluation
AWS Bedrock

[geospatial]

GDAL · ODA
Mapbox · GeoJSON
Coordinate systems
CAD: DWG / SHP
Cesium

[cloud & data]

AWS: Lambda, Step Functions, ECS
Azure · CI/CD
Docker
PostgreSQL · SQL Server
MongoDB

$ cat about.md

I studied Computer Science at IBA Karachi and started at Folio3 as a trainee, which is a polite way of saying I learned a large codebase the hard way. Somewhere between porting a C++ SDK into C# and rebuilding a pipeline as a state machine, I figured out what I actually like: problems where the fix is a better model of the system, not another layer on top of it.

I take ownership of problems whether or not they're strictly mine, I move fast, and I'd rather be honest about a weakness than dress up a number. I came back from a serious football injury — facial fracture, surgery, the whole recovery — and a rough patch after it, and that taught me setbacks are information, not verdicts.

What I want next: a team that builds things which work in the real world — in a paddock, not just a browser tab.

$ ls ~/beyond-code

Football, since I was eight

Captain and coach of the men's team at IBA — training, strategy, diet plans and trials for a 45-player squad with no coach, and a near-unbeaten season with a home Sports Fest title. I still play, and I coach and organise in Karachi's grassroots scene.

Retro & vintage motorbikes

Old machines, mechanical honesty, character you can hear. A genuine passion — the kind where I'll happily lose an afternoon to someone else's carburettor.

Mountains, and the gym in between

Travel and hiking reset me — somewhere new, on foot. I train regularly too; discipline off the screen keeps me sharp on it.

Giving back

Taught mathematics to underprivileged students at The Citizens Foundation, volunteered at community Iftar drives serving around 1,500 people, donate blood regularly, and organise need-based fundraising in the football community when a family hits a medical emergency.