Open to full-time AI/ML Engineering roles

Engineering applied AI from model weights to production systems.

BS Artificial Intelligence graduate from FAST-NUCES. Specializing in computer vision pipelines (YOLOv11), sensor fusion (Kalman filters), autonomous agents, and low-latency FastAPI architectures.

PythonFastAPIYOLOv11RAGSupabaseFlutterHugging FaceStreamlit
Loading 3D Visualizer...
Engineering Philosophy

Building systems where the machine learning model is only one piece of the puzzle.

I am a BS Artificial Intelligence graduate from FAST-NUCES with a deep commitment to practical, deployable AI engineering. While training models in Jupyter notebooks is foundational, true engineering begins when that model must interact with raw, noisy physical signals, high-throughput microservices, and human operators.

My work spans from edge computer vision (YOLOv11 custom architectures) and mathematical sensor fusion (Kalman filters) to autonomous agent orchestration and retrieval-augmented assistance. I believe in engineering complete feedback loops: from data curation and fine-tuning (LoRA/PEFT) to containerized inference pipelines with sub-100ms latency guarantees.

Portfolio Showcase

Featured AI & Systems Engineering

End-to-end production systems covering computer vision, autonomous agents, sensor fusion, and high-performance backends.

Computer VisionFlagship System

OmniDrive AI

Intelligent Automotive Diagnostic & Edge Perception Ecosystem

An automotive diagnostic and safety platform combining real-time computer vision, multi-modal OBD-II sensor data, 1D Kalman filter sensor fusion, and retrieval-augmented mechanical assistance.

PythonFastAPIYOLOv111D Kalman FilterSupabasePostgreSQLFlutterRAGpgvector
View Architecture & Code
99.1%
Top-1 Accuracy
50
Component Classes
92ms
Inference Latency
26,820
Annotated Frames

Pipeline Architecture

OBD-II CAN Bus< 5ms
1D Kalman Filter< 1ms
FastAPI Gateway12ms
YOLOv11 Vision Core92ms
pgvector RAG18ms
Flutter Client HUD60 FPS

Ecosystem Modules

CAN-Bus & Telemetry Ingestion
  • -Real-time OBD-II PID polling pipeline streaming diagnostic telemetry at 20Hz.
  • -1D Kalman filter sensor fusion denoising speed, acceleration, and RPM fluctuations.
  • -Engine anomaly detection flagging overheating and misfires prior to CEL triggers.
Edge Computer Vision Core
  • -Custom YOLOv11 model fine-tuned on 26,820 automotive engine and chassis images.
  • -Multi-label classification of 50 mechanical failure modes and wear indicators.
  • -FastAPI inference server optimized with batched tensor ingestion and asynchronous dispatch.
Retrieval-Augmented Diagnostic Assistant
  • -RAG pipeline indexed with pgvector cosine similarity over vehicle technical service bulletins.
  • -Context-grounded troubleshooting engine outputting verified step-by-step DIY repair workflows.
  • -Integration with vehicle diagnostic codes (DTC) for automated root-cause analysis.
Production Infrastructure & Client HUD
  • -Supabase PostgreSQL managing row-level security, auth, and relational telemetry history.
  • -Role-based ecosystem supporting vehicle owners, certified mechanics, and fleet managers.
  • -Cross-platform Flutter application rendering real-time telemetry HUD and AR diagnostics.
Interactive Architecture Inspector

OmniDrive Signal-Flow & Edge Processing Pipeline

Live Sensor Denoising (Kalman vs Raw)
Raw Noise Denoised State
Tensor InputScalar z_k (Noisy raw velocity/accel)
Tensor OutputScalar x_hat_k (Optimal minimum variance estimate)

Recursively denoises sensor noise and removes accelerometer drift using predictive covariance updates.

kalman.pyProduction Verified
class KalmanFilter1D:
    def __init__(self, q=1e-4, r=1e-2):
        self.q, self.r = q, r  # Process & Measurement variance
        self.x, self.p = 0.0, 1.0
    
    def update(self, z: float) -> float:
        # Predict & Correct cycle
        self.p += self.q
        k = self.p / (self.p + self.r)  # Kalman Gain
        self.x += k * (z - self.x)
        self.p *= (1.0 - k)
        return self.x
LLM & Agents

AI Job Application Agent

Autonomous Multi-Agent Pipeline with Human-in-the-Loop Review

Autonomous AI agent that orchestrates web scraping via Playwright, evaluates candidate-role alignment using Groq LLaMA-3 reasoning, completes multi-step application workflows, and logs submissions to Google Sheets.

PythonGroq / LLaMA-3PlaywrightGoogle Sheets APIPydantic
View Architecture & Code
4.2x
Speed Multiplier
98.5%
Field Extraction
0
Hallucinated Fields

Pipeline Architecture

Target Board
DOM Parser
LLaMA-3 Reasoning
Form Automation
Review Checkpoint
Live Audit Sheet

Key Engineering Highlights

  • Dynamic browser orchestration utilizing Playwright with human-behavior emulation.
  • Structured LLM reasoning ensuring exact schema conformance via Pydantic validators.
  • Human-in-the-loop validation checkpoint prior to irreversible application dispatch.
  • Real-time status synchronization and rate-limit handling across job boards.
LLM & Agents

Serene: AI Mental Health Companion

LoRA Fine-Tuned LLM with Local Emotion Classification

Mental wellness assistant fine-tuned with Low-Rank Adaptation (LoRA) on custom empathetic dialogue corpora, combined with a local DistilRoBERTa emotion classifier for real-time sentiment tracking.

PyTorchHugging FaceLoRA / PEFTDistilRoBERTaFastAPIStreamlit
View Architecture & Code
16-bit
Quantized Weights
7
Emotion Dimensions
< 45ms
Sentiment Latency

Pipeline Architecture

User Dialogue
DistilRoBERTa Classifier
Safety Gate
LoRA Fine-Tuned LM
Response Synthesis

Key Engineering Highlights

  • Fine-tuned causal language models using PEFT/LoRA to retain domain knowledge without catastrophic forgetting.
  • Integrated lightweight emotion classifier providing conversational guardrails and crisis detection.
  • Deterministic safety evaluation pipeline preventing ungrounded psychiatric recommendations.
  • Deployed with FastAPI inference backend and responsive Streamlit clinician dashboard.
LLM & Agents

NewsLens & TicketIQ: NLP Intelligence

Fine-Tuned BERT & Zero-Shot Classification Engine

Production-oriented NLP pipeline combining fine-tuned BERT for high-throughput news categorization with BART-large-MNLI for zero-shot customer support ticket classification and intent routing.

PythonBERTBART-large-MNLITransformersFastAPIDocker
View Architecture & Code
94.8%
BERT F1-Score
14
Topic Taxonomies
< 25ms
CPU Throughput

Pipeline Architecture

Raw Stream
Tokenization
BERT Classifier
Zero-Shot NLI
Webhook Routing

Key Engineering Highlights

  • Fine-tuned transformer encoder heads for multi-class hierarchical topic classification.
  • Zero-shot natural language inference pipeline for dynamic triage without re-training.
  • ONNX runtime model quantization delivering 3.2x CPU throughput acceleration.
  • Containerized microservice ready for Kubernetes horizontal pod auto-scaling.
Systems & Web

Protomotive Car Care Studio

High-Performance Interactive Digital Showcase

High-end detailing studio showcase engineered with Next.js App Router, GSAP timeline choreography, Lenis inertial smooth scrolling, and Tailwind CSS v4.

Next.js 15+Tailwind CSS v4GSAPLenisTypeScript
View Architecture & Code
100
Lighthouse Performance
60 FPS
Animation Budget
< 0.05
CLS Score

Pipeline Architecture

Next.js SSR
Lenis Smooth Scroll
GSAP Timelines
Fluid Layout

Key Engineering Highlights

  • Inertial smooth scrolling architecture integrated with GSAP ScrollTrigger.
  • Server-rendered Next.js architecture with zero layout shift (CLS < 0.05).
  • Micro-interactions and fluid viewport-aware visual typography.

Research Prototypes & Client Systems

Computer Vision

Construction Helmet Detection

Edge computer vision model trained on YOLOv5 for construction safety PPE compliance.

Computer Vision

Mesh Shift Visualizer

Interactive WebGL visualization for complex 3D mesh topological coordinate transforms.

Sensor Fusion & Edge

Clinical Heart Disease Prediction

Supervised classification pipeline benchmarked across Random Forest and XGBoost ensembles.

Systems & Web

Verto Digital Agency Platform

High-conversion digital growth agency platform with dynamic Framer Motion interactions.

Repository
Background & Trajectory

Engineering Experience

Professional engineering internships and academic teaching appointments at FAST-NUCES.

Apr 2026 – Jun 2026

AI/ML Engineering Intern

DevelopersHub CorporationRemote / Pakistan
  • Designed and deployed applied AI systems focusing on LLMs, RAG pipelines, and NLP classification.
  • Engineered modular applications integrating Hugging Face models, FAISS vector stores, and Streamlit interfaces.
  • Implemented fine-tuning pipelines using LoRA for custom instruction-following tasks.
Jun 2025 – Aug 2025

AI Developer Intern

NexiumLahore, Pakistan
  • Built full-stack applications using React, Next.js, and Supabase.
  • Integrated the Gemini API and automated LLM workflows using n8n orchestration.
  • Developed data pipelines for web scraping, translation, and structured data generation.
Sep 2024 – Jun 2025

Teaching Assistant — Programming Fundamentals

FAST-NUCESLahore, Pakistan
  • Mentored students in C++ and core programming fundamentals.
  • Evaluated code quality, logic-building, and project implementation.
Capabilities & Tools

Technical Toolkit

Core frameworks and technologies used across machine learning research, inference backends, and frontend delivery.

AI / ML

  • Python
  • PyTorch
  • Scikit-learn
  • Computer Vision
  • NLP
  • LLMs
  • RAG
  • LoRA

Models & Retrieval

  • YOLOv11
  • BERT
  • GPT-Neo
  • DistilRoBERTa
  • BART-large-MNLI
  • FAISS
  • BM25
  • SentenceTransformers

Backend & Data

  • FastAPI
  • REST APIs
  • PostgreSQL
  • Supabase
  • Firebase
  • Node.js

Frontend

  • Flutter
  • React
  • Next.js
  • Streamlit

Infrastructure & Tools

  • Docker
  • Git
  • GitHub Actions
  • Vercel
  • n8n
  • Postman

Verified Credentials

Formal industry internship certifications in AI/ML engineering and production systems.

Apr 2026 – Jun 2026

AI/ML Engineering Internship Certificate

DevelopersHub Corporation

6-week AI/ML Engineering internship.

Jun 2025 – Aug 2025

AI Developer Internship Certificate

Nexium

Full-stack AI application development using React, Next.js, and Gemini API.