Nimra Khalid — AI Engineer

About Me

Who I Am

An AI Engineer focused on turning research into production-grade systems. I design and ship intelligent applications — from autonomous agent frameworks and multimodal data pipelines to computer vision safety systems and semantic search engines.

I hold a BS in Artificial Intelligence from SZABIST Islamabad (CGPA 3.68/4.0, Silver Medalist) and have shipped real AI in professional settings at RCM Nexus, Biome, and Disruptive AI.

Certified as a Microsoft Learn Student Ambassador, I bridge cutting-edge research with actual engineering. My philosophy: I build AI that ships, not just demos.

Currently seeking exciting AI Engineering opportunities.

📧 Say Hello ⌨️ GitHub

3.68

CGPA / 4.0

🥈

Silver Medalist

3+

AI Roles

4

Certifications

Capabilities

My Tech Stack

Tools and technologies I use to build intelligent, production-ready AI systems.

🧠

01

Agentic & LLM

Agent Loops LLaMA LLaVA Prompt Eng. Tool Registries

🔬

02

ML / DL Frameworks

TensorFlow88%

Deep Learning85%

Computer Vision82%

NLP85%

🌐

03

Tools & Interfaces

Streamlit FAISS OpenCV Laravel MySQL

💻

04

Languages

Python95%

MySQL78%

Laravel / PHP70%

📊

05

Data & Pipelines

Data Pipelines88%

API Integration90%

Multimodal ML82%

☁️

06

Deployment

Netlify Git / GitHub Streamlit Cloud REST APIs

My Work

Featured Projects

AI systems built from idea to production deployment.

✒️

🟢 Live Demo View Repo ↗

nanoglyph — Transformer Built From Scratch

A decoder-only transformer with hand-written causal self-attention — no nn.MultiheadAttention, every matmul and mask is explicit. Trained character-level on Shakespeare's complete works (loss 4.56 → 1.61 over 2000 steps), with a test proving zero future-token leakage rather than just checking output shapes.

PyTorch Transformers Streamlit pytest

🎯

View on GitHub ↗

HireSmart — AI Recruitment System

End-to-end recruitment platform — job postings, GPT-4 resume screening against the job description, interview panel scheduling and evaluation, personality assessment, and offer letters, with role-based dashboards.

Laravel GPT-4 Python FYP

🗂️

Private Repo

TrainSet AI — Multimodal Dataset Builder

Unified data ingestion pipeline processing audio, images, PDFs, and text to generate structured ML training datasets. Uses LLaMA and LLaVA to convert speech and visual content into clean annotated data.

LLaMA LLaVA Multimodal Python

🔍

Private Repo

Semantic Search Engine (LLM-Free)

Fast document retrieval using embedding-based similarity search without large language models. Document chunking and cosine similarity with FAISS for accurate information retrieval via Streamlit.

FAISS Embeddings Streamlit Python

📄

View on GitHub ↗

DocIQ — RAG Document Intelligence

Retrieval-augmented document Q&A — FastAPI backend, ChromaDB vector store, Claude for generation. Ships with a retrieval-quality eval harness (hit@3: 90%) that gates CI, and 28 tests exercising real embedding-based retrieval rather than mocks.

RAG FastAPI ChromaDB CI/CD

🔎

View on GitHub ↗

groundtruth — Research Agent

Plans a question into sub-questions, researches each with live web search, and synthesizes one answer where every claim cites the source that actually supports it — pulled from real search results, never invented.

Agentic AI Claude API Python pytest

🤖

🟢 Live Demo View Repo ↗

agent-pipeline — Multi-Agent Pipeline

Researcher, writer, and critic agents hand off work in sequence with live streaming — the critic checks every claim in the draft actually traces back to the research brief and triggers a real revision when it doesn't.

Next.js TypeScript Tailwind Groq

🏥

Private Repo

NPI Physician Data Extraction System

Python-based large-scale data extraction system using the CMS NPI Registry API to collect physician records. Structured validation, retry mechanisms, and request throttling for production reliability.

Python API Data Pipeline Automation

Work History

My Experience

Where I've worked and what I've built along the way.

Jul – Dec
2025

AI Engineer

RCM Nexus · Remote

Developed a Python-based data extraction system using the CMS NPI Registry API to collect and process large-scale physician records including identifiers, specialties, and licensing.
Engineered structured validation, retry mechanisms, and request throttling to ensure data consistency and reliability at scale.
Reduced manual data collection effort significantly through end-to-end pipeline automation.

Jul – Oct
2025

AI / ML Engineer

Biome · Remote

Designed a custom agent-loop framework enabling LLMs to reason autonomously, select tools, and execute multi-step tasks without human intervention.
Built modular system components: tool registries, prompt orchestration layers, and observation handlers.
Contributed to a production-ready agentic AI system used in client-facing automation workflows.

Jun – Sep
2024

AI Engineer Intern

Disruptive AI · NIC Islamabad

Developed TrainSet AI — a multimodal data ingestion platform processing audio, images, PDFs, and text into structured training datasets using LLaMA and LLaVA.
Contributed to a computer-vision-based construction site safety system using object detection and segmentation to monitor PPE compliance in real time.
Collaborated on model fine-tuning, dataset curation, and deployment pipelines.

May
2025

Bachelor of Science — Artificial Intelligence

SZABIST Islamabad · Pakistan

CGPA: 3.68 / 4.0 — Silver Medalist (Top Graduating Student)
Coursework: Machine Learning, Deep Learning, NLP, Computer Vision, Data Structures & Algorithms.

Credentials

Certifications & Awards

🏆

Microsoft Learn Student Ambassador (MLSA)

Microsoft · SZABIST

⚙️

Agentic AI Engineering

Professional Certification

🤖

Generative AI For Everyone

DeepLearning.AI

📐

Machine Learning Specialization

Andrew Ng · Coursera

🥈

Silver Medal — Top Graduate

SZABIST Islamabad · 2025

Open Source

GitHub Repositories

Public projects and experiments on GitHub.

🩺

🟢 Live Demo View Repo ↗

Diabetes Risk Predictor

Three models compared honestly on cross-validated ROC-AUC (logistic regression won, beating gradient boosting and the original ANN), with SHAP explaining each individual prediction — not just a number. Live, working demo, not just code.

scikit-learn SHAP Streamlit CI/CD

🍎

View Repo ↗

Fruit Detection Using YOLOv8

Real-time object detection system using YOLOv8 to identify and classify fruits. Showcases computer vision pipeline from dataset preparation to model inference.

YOLOv8 Computer Vision Python

💵

View Repo ↗

KNN Banknote Authentication

K-Nearest Neighbors classifier for authenticating banknotes using wavelet-transformed image features. Clean implementation with accuracy benchmarking.

KNN ML Jupyter

🌳

View Repo ↗

Machine Learning Decision Tree

Decision Tree classifier implementation from scratch and with sklearn, with visualizations and comparison of tree depths, pruning strategies, and performance metrics.

Decision Tree ML Jupyter

🤖

View Repo ↗

RL Agent in Grid World

Reinforcement learning agent navigating a grid world environment using Q-learning. Demonstrates core RL concepts: state space, reward shaping, policy learning, and convergence.

Reinforcement Learning Python

♛

View Repo ↗

N-Queens Genetic Algorithm

Solving the classic N-Queens combinatorial problem using a Genetic Algorithm — selection, crossover, and mutation operators tuned for fast convergence.

Genetic Algorithm AI Jupyter

⌨️ View All Repositories on GitHub

Community

Open Source Contributions

Real pull requests to other people's projects, not just my own repos.

🌈

View Repo ↗

chroma-core/chroma

Three PRs against the vector database I used to build DocIQ: enforcing absolute imports via flake8-tidy-imports (#7507), and fixing two Jest tests whose try/catch pattern let them pass vacuously even when the code under test didn't throw (#7508, #7509).

Python TypeScript Open PR ×3

📖

View Repo ↗

Unstructured-IO/unstructured

PR #4420: the xy-cut layout sorting used for PDF partitioning always read blocks left-to-right, which is the wrong order for right-to-left scripts like Arabic and Hebrew. Added an opt-in rtl mode with tests proving the existing left-to-right behavior stays unchanged by default.

Python Document Parsing Open PR

🍳

View Repo ↗

anthropics/claude-cookbooks

PR #790: a new cookbook notebook contributed to Anthropic's official Claude examples repository.

Claude API Open PR

Get In Touch

Let's Connect

Ready to build something intelligent? I'm actively seeking exciting AI Engineering opportunities. Whether you're building an LLM product, a computer vision system, or an agentic AI platform — let's talk.

✉️ nimra3261@gmail.com ⌨️ github.com/Nimra3261 🔗 LinkedIn 📞 +92 333-5100-136

📧 Email Directly