system v4.0 · cyberpunk runtime 0%
Available · Open to AI Engineer roles

Muhammad Fahad.

> AI Engineer

I build production AI — RAG systems, agentic pipelines, and LLM applications that survive the jump from notebook to server. Currently shipping at Veeivs & Provoxio. Based in Islamabad.

▶ See the Work ↓ Download CV ✉ Contact
// Find me
AI / ML // SYSTEMS
LLMs
RAG
Agentic
PyTorch
Overall
0+
AI systems in production
Overall
0mo+
Production experience
DocuMind · Veeivs
0%
System uptime
DocuMind · Veeivs
<2s
Median response latency
Speak-Align · Provoxio
0%
QA time automated
Speak-Align · Provoxio
0/day
Calls scored daily
Meeting Assistant
0min
Recap per 60-min call
SupportIQ
0 stages
AI pipeline depth
// 01 — About

Who is Fahad?

An AI Engineer obsessed with the messy middle — where a model that works on a laptop has to survive real traffic at 3 AM.

I'm Muhammad Fahad, an AI Engineer based in Islamabad, currently building production systems for Veeivs and Provoxio. My work lives between two worlds: the engineering side — Python services, Docker, AWS, Nginx, queues — and the AI side — RAG pipelines, agentic workflows, LLM evaluation, speech systems.

Most AI projects break between the notebook and the server. I focus on the part that survives: grounded retrieval, recoverable agents, and pipelines that don't silently drop traffic.

"Most AI work fails between the notebook and the server.
I do the part that survives the server." — Working principle
i
01
RAG & Retrieval
Vector search · Qdrant · embeddings · hybrid retrieval
ii
02
Agentic AI
LangChain · n8n · multi-step recovery · tool use
iii
03
Production Deployment
Docker · Nginx · HTTPS · monitoring · failure surfacing to Slack
iv
04
Speech AI
Whisper · diarisation · sentiment
v
05
Cloud (AWS)
EC2 · S3 · Lambda · cost-aware design
// 02 — Selected Work

Production Systems

Five shipped systems, real metrics, real production. Click ▶ Run Demo on any card to see a simulated interaction with the live system.

01
Live · Provoxio

Speak-Align

// Call-center QA automation · agentic pipeline

Ingests 50–100 call recordings daily, transcribes them with Whisper, scores each call against a defined script, and flags order violations and agent deviations before a human listens to a second of audio. Daily digest delivered to ops by 9 AM.

~55%
QA time saved
100/d
Calls evaluated
14
Scoring rules
WhisperLangChainn8n FastAPIPostgres
speak-align@provoxio:~$
02
Live · Veeivs

DocuMind

// Source-grounded document Q&A · RAG

Production RAG chatbot deployed on bare Ubuntu 24 with Docker Compose, Nginx, HTTPS, and a custom domain. Answers cite the exact source chunk so the team can verify, not vibe-check. Per-tenant isolation; hot-swappable embedding models.

< 2s
Median latency
10K+
Docs / corpus
99.4%
Uptime
StreamlitFlowiseQdrant OllamaGroqDocker
documind@veeivs:~$
03
Production

Meeting Assistant

// Recording → recap → tasks · end-to-end

Watches a Google Meet folder, transcribes each new recording, extracts summary, decisions and action items, e-mails the recap to attendees, and opens prioritised tasks in ClickUp — without a human touching a button. Failure modes surface in private Slack.

3 min
Turnaround / 60-min call
100%
Auto-delivery rate
0
Silent drops
n8nGroq WhisperCloudinary Gmail APIClickUp API
meeting-asst@veeivs:~$
04
Open Source

Mock Interview

// Speech-aware interview rehearsal · GenAI

Interactive AI mock-interview app: generates role-specific questions, records and transcribes spoken answers, scores sentiment and confidence with transformer LLMs, and exports structured JSON feedback reports. Built originally as a teaching tool for juniors.

JSON
Report export
15+
Question templates
3
Scoring axes
PythonStreamlitGemini API Transformers
mock-interview:~$
05
Live · AWS EC2

SupportIQ

// AI customer support agent · RAG + routing + ticketing

Production-style AI support agent that classifies customer intent and sentiment, retrieves policy answers via RAG, runs order lookups, auto-creates tickets, escalates urgent cases, and generates human-agent handoff summaries — deployed on AWS EC2 behind Nginx.

8
AI pipelines
RAG
Qdrant + Ollama
3
Docker services
GroqQdrantOllama StreamlitDockerNginxAWS EC2
↗ Live { } GitHub
supportiq@aws:~$
// 03 — Journey

The Trajectory

A focused journey building real-world AI systems while sharpening the engineering fundamentals underneath.

Now Dec 2025 — Present

AI Engineer · Onsite

// Veeivs
Shipped DocuMind to production on Ubuntu 24 (Docker, Nginx, HTTPS). Built the Meeting Assistant pipeline end-to-end. Designed a speaker-separation & transcription service. Automated core business workflows with agentic bots.
Now Nov 2025 — Present

AI Developer · Project-based

// Provoxio Ltd · Remote
Engineered Speak-Align: processing 50–100 call recordings daily to evaluate script compliance and flag agent deviations — cutting manual QA review time by ~55%.
Cert. 2024

AWS Solutions Architect · Associate

// Amazon Web Services
Cloud architecture on AWS — designing scalable, resilient, and cost-optimised systems. Applied directly to deployment work at Veeivs.
Cert. 2024

Certified AI Engineer · Associate

// School of AI · Udemy
End-to-end AI engineering: model development, deployment, and applied machine learning. Reinforced production discipline.
Edu. 2023 — 2027

BSc · Computer Science

// NUML · Main Campus, Islamabad
Foundations: algorithms, data structures, software engineering, applied AI/ML. All production work above runs alongside the degree.
// 04 — Stack

The Toolkit

The stack I actually use in shipped systems — not the wishlist version.

04 / a

AI / ML core

  • Large Language ModelsDaily
  • RAG & vector retrievalDaily
  • LangChainDaily
  • Agentic AIOften
  • Prompt EngineeringDaily
  • Transformers · PyTorchOften
  • Whisper · Speech-to-TextOften
04 / b

Engineering runtime

  • PythonDaily
  • FastAPIDaily
  • Docker · ComposeDaily
  • Nginx · HTTPSOften
  • REST APIsDaily
  • Git · GitHubDaily
  • Linux · Ubuntu 24Daily
04 / c

Cloud & data plumbing

  • AWS EC2Often
  • AWS S3Often
  • AWS LambdaOften
  • Qdrant (vector DB)Daily
  • n8n (orchestration)Daily
  • FlowiseOften
  • StreamlitOften
04 / d — Proficiency radar
Production proficiency
Comfort / familiarity

Where I actually ship.

Headline-grade skill lists everywhere are flattering and useless. This radar shows where my output is real — not where I've read a tutorial. Toggle between production proficiency (have I shipped this to a paying user?) and comfort (could I be productive in this on day one?).

// 05 — Live Console

Try the Systems

Most portfolios show screenshots. Pick a command below, or type your own (try help) to see a simulated interaction with one of my live systems.

fahad@portfolio:~$ — interactive
$ welcome --help
→ pick a command below, or type one and hit enter
→ try: help, whoami, ls, skills, contact, clear, sudo make-me-a-sandwich
$
// 06 — Signal

What people say

A few words from the operators, founders, and teammates I've shipped alongside. Names redacted on request where required.

★★★★★

"Fahad doesn't ship demos — he ships systems. Our RAG pipeline went from 'cool prototype' to handling real customer questions in production in about two weeks. He profiles, he benchmarks, he documents. Rare combination."

H
Hassan Ali
Operations Lead — Provoxio
★★★★★

"He thinks like a builder, not just a coder. When the Whisper pipeline started timing out under load, he didn't reach for a bigger box — he re-architected the queue, added back-pressure, and dropped p95 latency by more than half. I learn things every time we pair."

M
Maria K.
Senior AI Engineer — Veeivs
★★★★★

"I came in barely knowing what 'embeddings' meant. Three months later I deployed my first agentic system because Fahad teaches the way a senior engineer wishes someone had taught them — concepts first, then code, then why the code is the way it is."

A
Ali R.
Junior ML Engineer (mentee)
// 07 — FAQ

Frequently asked

The questions I get most from recruiters, founders, and engineers reaching out cold. Click any to expand.

Yes — I'm currently open to full-time AI engineering roles (remote-first, EU/US/MENA time zones work fine from Islamabad) and to select contract work for serious RAG, agentic-systems, or speech-AI builds. I'm at Veeivs and Provoxio in part-time/contract capacity, so I can ramp quickly without untangling a long notice period.
It means I've built systems that survive real users — not just a Jupyter notebook with a 5-document corpus. That includes: chunking strategies that respect document structure, hybrid retrieval (dense + sparse), re-ranking, citation-grounding so the model can't hallucinate sources, latency budgets, observability (token cost per query, retrieval recall, end-to-end p50/p95), and graceful degradation when the LLM API rate-limits you mid-conversation. DocuMind at Veeivs runs this stack today.
Comfortable, not specialist. I containerize my own services (Docker), wire reverse proxies (Nginx), set up systemd units, and deploy to AWS EC2 / Lightsail. For anything beyond that — multi-region, complex IAM, full IaC — I work alongside a dedicated DevOps engineer. I'm AWS Cloud Practitioner certified and currently working toward AWS AI Practitioner, so my fluency on the cloud side is increasing every month.
I'm based in Islamabad (PKT / UTC+5). I overlap comfortably with EU mornings, MENA full day, and US-East mornings. For US-West teams I can shift my schedule a few hours and have done so for two engagements already. I default to async-first communication, daily written updates, and showing up for sync calls when they actually need to be sync.
Anything where the AI is doing real work, not generating slop. Agentic systems that automate genuinely tedious workflows. RAG over corpora that humans physically cannot read in a lifetime. Speech systems that help people communicate better. I get the most out of teams that ship to real users, take latency and cost seriously, and treat the model as one component of a larger system — not as a magic box.
Honest answer: if you need a 10-year veteran who's run an ML org, hire them. What I bring is a different shape — I'm deep in the current generation of tooling (LangChain/LangGraph, vector DBs, modern agentic patterns) in a way that engineers trained on pre-2023 ML often aren't, I move fast, I write tight code, and I'm at the career stage where I'm still doing the work myself rather than delegating it. For a startup or a focused AI team inside a larger org, that's often the right trade.
// 08 — End of transmission

Let's build
something real.

If you've got a problem that genuinely needs AI — not a hype-driven sprinkle of GPT — I'd love to hear about it. The fastest channel is email; the form below builds a draft for you.

Email
muhammad.fahad1213@gmail.com
LinkedIn
linkedin.com/in/fahad2703
GitHub
github.com/M-Fahad27
Phone
+92 349 532 6867

Or send a structured ping.

Fill this in and I'll get an email with everything pre-formatted — no copy-pasting from a generic contact form. Best for project inquiries with a defined scope and a budget range.

  • Response within 24 hours on weekdays
  • Free 30-minute scoping call for serious inquiries
  • NDAs welcomed before any specifics
⌘ esc
↑↓ navigate ↵ select esc close
F
Fahad AI
Online · Avg reply ~2s
Hey 👋 I'm a small assistant trained on Fahad's portfolio. Ask me about his stack, projects, availability, or how to get in touch.

Keyboard shortcuts

K or ⌘K Open command palette
? Show this shortcuts panel
T Cycle visual theme
S Toggle UI sound
M Toggle matrix rain effect
G Scroll back to top
1—5 Jump to section (about / projects / timeline / stack / contact)
Esc Close any open overlay
↑↑↓↓←→←→BA ??? — try it
✓
Achievement
unlocked
ACCESS GRANTED
// cyberdeck unlocked · welcome, operator