AI Engineer Islamabad · Pakistan

Evidence first. Systems second. Hype never.

Building intelligent systems that turn scattered signals into clear action.

I build production-grade AI systems, enterprise RAG architectures, and custom automation engines that convert raw public data into structured, decision-ready intelligence.

Start a conversation View GitHub

For teams building intelligence, automation, or AI-enabled products.

10+ Public Repositories
Production AI Workflows Enterprise RAG & Agentic APIs
Product-Minded Engineering Tech complexity mapped to business value

What I build Intelligent systems that automate research, analysis, and repetitive work.

Operating style Understand the problem. Build the simplest reliable solution. Improve it.

Positioning

More than an AI engineer. I build systems with product instincts.

My work sits where automation, intelligence, and decision-making meet. The pattern across my strongest projects is consistent: gather noisy public information, structure it carefully, add reasoning only where it helps, and ship the result as something operational.

That pattern shows up across competitive intelligence, knowledge systems, B2B lead generation and outreach, and applied ML work. It also shapes how I think about products: trust signals, cognitive fluency, deployment discipline, and interfaces that make complexity feel manageable.

My work

Projects that best represent how I build.

These are the projects that best show how I think: evidence handling, automation depth, applied AI, and systems architecture that can survive real use.

01 Flagship platform

Competitive Intelligence Engine

I built this to turn public web data into source-cited intelligence, decision-ready reports, and structured business evidence.

Problem: market research and intelligence gathering is slow, manual, and prone to unverified claims.
Solution: an automated pipeline that ingests, normalizes, and validates web sources, outputting citation-backed reports.
Impact: turns noisy web research into traceable evidence a team can act on.

PythonFastAPIPostgresDockerOpenRouter

Open repository

02 Knowledge system

WhatsAppRAG

I designed this as a searchable second brain with semantic search, full-text search, and automated ingestion.

Problem: unstructured communications hold vital knowledge that gets rapidly buried and forgotten.
Solution: a searchable personal brain integrating vector embeddings, full-text search, and automated ingestion.
Impact: makes dispersed knowledge searchable, reusable, and operational.

PythonTypeScriptSQLiteChromaDBStreamlit

Open repository

03 Data + OSINT

B2B Lead Gen and Outreach

I turned B2B research workflows into a web-ready intelligence stack for discovering and enriching company data at scale.

Problem: scalable company data extraction and enrichment is often limited by script-based barriers.
Solution: a production-grade SaaS architecture leveraging distributed Celery workers, caching, and rate-limited APIs.
Impact: moves B2B intelligence from one-off scripts toward a repeatable service.

FastAPIMongoDBRedisCeleryDocker

Open repository

04 Applied AI media

Story2Audio

I built an emotion-aware text-to-speech service combining Bark generation, emotion detection, and async serving.

Problem: standard text-to-speech systems produce flat, robotic audio that lacks context-aware emotion.
Solution: an emotion-aware synthesis pipeline that uses NLP classifier prompts to direct raw audio generation.
Impact: connects model experimentation to a usable, async product surface.

PythonBarkRoBERTagRPCGradio

Open repository

05 Applied ML research

M2N2 Model Fusion

I explored a hybrid vision setup using EfficientNet-Lite0, a tiny recursive model, and weight fusion.

Problem: standard models can leave performance on the table.
Solution: test a hybrid architecture with reproducible configs and fusion logic.
Impact: demonstrates disciplined experimentation below the application layer.

PythonComputer VisionPyTorchCIFAR-10

Open repository

06 Security demo

Post-Quantum Flask App

I built this as a compact demo of hybrid encryption with Kyber512 and AES-GCM.

Problem: advanced cryptography often feels inaccessible.
Solution: expose key generation and encryption through a simple interface.
Impact: makes advanced cryptography easier to inspect through a compact interface.

PythonFlaskKyber512AES-GCM

Open repository

Approach

How I tend to think when the work is real.

Evidence before narrative

Facts get normalized, cited, and verified before intelligence becomes opinion. That habit shows up in both product design and backend architecture.

Automation with boundaries

The best systems do heavy lifting without becoming reckless. Clear limits, guarded workflows, and observable failure modes matter.

Product instinct inside engineering

Technical systems should also reduce friction, communicate trust, and make decisions easier for the person on the other side.

Structured execution

Ship, verify, audit, refine. The process is methodical because reliability matters more than looking clever in the first pass.

Stack

Built around practical AI engineering.

My workflow is Python-heavy and API-first, with strong comfort around automation, data extraction, retrieval, and production deployment.

Core

Python
FastAPI
Docker / Compose
Git

AI + data

RAG systems
LLM integrations
Scraping / OSINT
MLOps workflows

Operating style

Evidence-backed reasoning
Multi-agent orchestration
Deployment analysis
Conversion-aware product thinking

AI tooling

Claude Code
OpenAI Codex
Hermes Agent
MCP + Google ADK

Contact

Have a problem worth structuring?

I do my best work on systems that need intelligence, automation, and grounded product thinking — especially when the input is messy and the output needs to be trustworthy.

qasimjl469@gmail.com +92 321 8648706 LinkedIn GitHub