Skip to content
Amaan Nadeem
KHI · GMT+5Open to senior roles
KHI · GMT+5

Open to senior roles

Amaan Nadeem, smiling, arms crossed, in a dark blazer and white shirt

Illustrative point-cloud scan of a living room, 4.21 metres wide, with Amaan standing in it.

Senior Full-Stack & AI Engineer — Karachi

The demo is the easy part. I build the part that holds up in production.

I buildAI agents, LiDAR apps, event-driven platforms, payment engines.

See selected work
7 yrs
shipping production software
100%
Job Success on Upwork, Top Rated
5.0★
average from 27 client ratings
2.8K
contributions in the last year
60+
production repositories
Now
— Senior Software Engineer at BetterFutureLabs (US, remote), since Aug 2025.

01 — Pull back

From one room to the open road.

  1. 00 · One room

    A room, measured.

    Wave Scan measures a water-damaged room in about forty seconds with LiDAR — and on phones without a depth sensor, the technician aims at each corner from two places and I triangulate it.

    Read the Wave Scan case study
  2. 01 · Cross-border

    Freight across borders.

    TIRGPT books, dispatches and tracks road freight across borders. The driver app streams background GPS to AWS IoT Core — and buffers offline when a border crossing drops the signal.

    Routes illustrative.

    Read the TIRGPT case study
  3. 02 · Systems

    Then the systems around them.

    Agents, marketplaces, payment engines, scrapers. Each one a set of parts that has to keep working once real people depend on it.

    See selected work

02 — Selected work · 6 systems

Things that held up.

iOS · LiDAR · Swift

Wave Scan — iPhone room measurement

An iPhone app that measures a water-damaged room in about forty seconds.

assumed room dimensions became measured ones, with provenance

~40 s
per room, LiDAR capture
239 / 426
non-merge commits, top contributor
~2,290
tests across the Swift package
  • Swift 6
  • SwiftUI
  • ARKit
  • RoomPlan
  • Euclid
  • TUSKit
capture_session · room 01● LiDAR
4.21 m~40 s per room

AI agent · Python · Bedrock

ClaimsAI — Discovery Agent V2

Contractors chat their way to a priced, insurance-grade estimate.

a 99,000-character prompt became a code state machine + structured extraction

256 s to 12.7 s
worst-case turn latency
851
tests in the V2 suite
~11k
lines written solo for V2
#1
committer
  • Python
  • FastAPI
  • Pydantic
  • Supabase
  • AWS Bedrock (Claude)
  • Stripe
discovery_agent / state_machine.py● turn 14
INTAKELOSS_TYPEROOMSMEASURELINE_ITEMSPRICEREVIEWESTIMATEre-ask ifincompleteCLAUDE · EXTRACTION (BEDROCK){ "room": "Kitchen", "dims_ft": [14, 13], "flood_cut_ft": 2, "confidence": "stated" }validated → state advancesCODE OWNS THE FLOW.THE MODEL ONLY EXTRACTS.replay harness ▸ real conversations
CONTRACTOR

Kitchen's about 14 by 13. Drywall's wet up to 2 ft.

AGENT

Got it — Kitchen 14×13 ft, 2 ft flood cut. Does the water reach the hallway?

illustrative turn

AI testing · Python · Playwright

Qove — AI-generated tests from real sessions

Real user sessions become AI-written tests that replay themselves.

reasoning on every run became reason once, replay many

$5.40 to $0.50
cost of one test run
953k to 163
error rows collapsed to distinct errors
1,100+
commits
  • Python
  • FastAPI
  • Postgres
  • AWS Bedrock
  • Playwright
  • Docker
qove / runnerreason once · replay many
  1. 01SDK~5 KB, batched events
  2. 02IngestFastAPI → Postgres
  3. 03Journeyagent-readable steps
  4. 04Testwritten once by AI
  5. 05Replayevery change
  6. 06Triageautonomous
reason on every run$5.40
reason once, replay$0.50

cost of one test run

error rows953k
distinct errors163

failure triage

Marketplace · Node · Elasticsearch

AccountMakers — AI sales-talent marketplace

A marketplace that matches employers with vetted sales talent.

~20 s talent search on MongoDB became under 2 s on Elasticsearch

~20 s to <2 s
talent search
455
API handlers
36
Lambda functions
  • TypeScript
  • Node.js
  • MongoDB
  • Elasticsearch
  • AWS Lambda
  • Next.js
talent search · p99MongoDB → Elasticsearch
MongoDB~20 s
Elasticsearch<2 s

talent search (target)

  1. 01Bullhorn ATSOAuth sync
  2. 02Elasticsearchtalent index
  3. 03Filterhard requirements
  4. 04Scorerules + embeddings
  5. 05Re-rankLLM, cost-gated
  6. 06Hireinterviews · offers

Freight · event-driven · AWS

TIRGPT — Cross-border freight platform

Freight booked, dispatched and tracked across borders.

freight coordinated by phone became a 24-state order lifecycle

24
states in the order lifecycle
1,000
GPS points buffered offline
41
permissions across 7 roles
  • Node.js
  • AWS EventBridge
  • SQS
  • DynamoDB
  • ECS Fargate
  • React Native
driver app · live GPS
TIRGPT cover — driver app with live GPS
ORDER LIFECYCLE24 states · enforced transitions

Health · Next.js · Claude

VitalLog — Personal health almanac

Blood tests, food and training in one place, read by an AI that asks before it writes.

lab PDFs nobody reopens became biomarkers charted over time

~6
days to build
124
tests
  • Next.js
  • FastAPI
  • Supabase (RLS)
  • AWS Bedrock (Claude)
  • USDA FNDDS
vitallog · biomarkers
VitalLog — your whole health, synthesized

More work → Convolytica · Stadiumatch · Events Scraper · System Atlas · Spendwise · JobPilot · Tastely · Learn to Earn

03 — More work

Smaller systems, same standard.

04 — Experience

Nearest first. Everything else recedes.

Seven years, read as a depth timeline: the current role on the near plane, earlier work falling back toward the vanishing point.

  1. 2025 — nowBetterFutureLabs · Senior Software EngineerUS, remote · Qove, ClaimsAI, Wave Scan
  2. 2022 — nowClient platforms · Founder-level engineer & technical leadAccountMakers, TIRGPT, Stadiumatch
  3. 2021 — 22Convolytica · Software EngineerSocial-listening NLP pipeline
  4. 2020 — 22Va8ive Digital · MERN & JAMstack DeveloperLed a small team · Best Employee of the Month
  5. 2020 — nowUpwork · Freelance engineerTop Rated · 100% Job Success · 5.0 from 27 clients
  6. 2019PIAIC · MERN Stack DeveloperStudent portal used by thousands of students
  7. 2022BS Computer Science · Karachi University (UBIT)

05 — Lab

Small things, built to be played with.

Nothing autoplays. Live pieces open in a new tab; the ones marked “on the bench” are still being built.

Ask my portfolio

from the case studies · cites sources

Pick a question, or type your own. Every answer cites the case-study section it came from.

Answers only from this site’s case studies · runs in your browser · no network calls

How it answers

Every answer is written from the case studies on this site and links to the section it came from. If the answer isn’t in them, it says “I don’t know” instead of guessing.

Runs entirely in your browser — no model, no network calls, nothing you type leaves the page.

Press ⌘K / Ctrl K anywhere.

06 — Words from clients

Words from clients

Upwork · verbatim

“Amaan worked as a leading programmer on our complex project which consisted from many microservices and complex event-driven flows between them. He implemented a lot of advanced features, he worked through all stages of the product from very early days to beta product.”

— Founder, TIRGPT · freight platform

“Amaan delivered outstanding work… demonstrating exceptional technical skills in AWS Cognito, DynamoDB integration, and production-ready architecture. His engineering maturity and attention to detail make him a valuable team asset.”

Product lead · Logistics platform · auth service

“What really stood out to me was his detailed explanation of his thought process… he didn't just work around issues, he pointed them out and provided a detailed plan on how to fix them.”

Product founder · Dashboard build

“With difficult coding tasks, Amaan is excellent and patient. Solves all the tasks and is a very hard worker.”

Client · 2020
07 · Contactopen to senior roles

Let’s build the part that holds up.

Open to senior full-stack and AI engineering roles — remote or relocation.

amaannadeem@outlook.com

Ask my portfolio