Fatima Sohail Shaukat

AI/ML Engineer & Researcher specializing in Healthcare AI, Physiological Signals, and Production Systems

Based in: Lahore, PakistanCurrently: AI Automation & Operations Analyst, Amazon E-CommerceExperience: 3 years

Fatima is an AI Engineer and researcher focused on building complex NLP, deep-learning, and automation systems for healthcare and enterprise applications. Her expertise spans multimodal physiological-signal modeling, retrieval-augmented generation, and efficient knowledge distillation for wearable, human-centered AI.

Experience

AI Automation & Operations Analyst

Amazon E-CommerceRemote, USAwork2026 – Present

Summary

Building AI workflows and automation for e-commerce operations, forecasting and reporting. Managing marketplace activity: product research, listing optimization, PPC campaigns and launches. Creating internal knowledge systems, documentation and workflows to support scalable growth.

What I did

Acted as the sole developer handling all automation work for a non-technical team.

AI workflowsAutomationForecastingPPC campaignsAmazon FBASeller CentralHelium 10

Lead AI Engineer

QuilvenRemote, UKworkSep 2025 – Aug 2026

Summary

Built and deployed an AI lead-sourcing and outreach system on multi-agent workflows (sourcing, enrichment, outreach), processing 3,oo0+ leads and replacing manual prospecting. Designed hybrid healthcare-data pipelines over the NPI Registry, PECOS and Playwright scraping, with EHR detection via patient-portal tracing. Developed ICP scoring and signal-based enrichment (Google Places, hiring data) to prioritize high-intent leads. Integrated HubSpot, Instantly and LinkedIn outreach through API and cloud workflows

What I did

Built and deployed an AI lead-sourcing and outreach system on multi-agent workflows (sourcing, enrichment, outreach), processing 3,000+ leads and replacing manual prospecting.
Built production-grade RAG and agentic systems.
Served as the solo developer, making all design and coding choices independently.

Results

Processed 3,000+ leads and replaced manual prospecting.

Multi-agent workflowsNPI RegistryPECOSPlaywrightEHR detectionGoogle PlacesHubSpotInstantlyLinkedIn APICloud workflows

Machine Learning Engineer

Meriti Inc. (LexisNexis)Remote, USAworkDec 2025 – Apr 2026

Summary

Worked on enterprise-scale AI drafting and retrieval systems for LexisNexis, focused on RAG reliability, citation grounding and agent-workflow validation. Built and maintained production PyTest infrastructure across 20+ modules with ~2,200 passing tests, high coverage, fast execution and fully mocked dependencies. Led a redesign of the citation architecture, replacing inline PID markers with structured reference anchors for precise source traceability. Implemented end-to-end handling of vault and uploaded documents across McP, drafting and grounding layers; built FastMcP validation clients for drafting pipelines and tool execution.

What I did

Developed production RAG and agentic systems as part of the company's internal systems.

Results

Built production PyTest infrastructure across 20+ modules with ~2,200 passing tests.

RAGPyTestMCPFastMcPLegalTechAgent-workflowsCitation grounding

Skills

Technical

Python

NLP

EEG Processing

Physiological-signal processing

Pandas

NumPy

LLMs

AI Agents

Knowledge Distillation

Model Compression

RAG

Multimodal Learning

PyTest

n8n

PyTorch

Supabase

XGBoost

OpenAI API

Git

CNNs

Time-series forecasting

Playwright

HubSpot

Instantly

LSTMs

Docker

LangChain

Hugging Face

Transformers

Computer Vision

OpenCV

OCR

YOLOv8

FastAPI

TypeScript

TensorFlow

Random Forest

Explainable AI

MCP

Django

PostgreSQL

React

Next.js

AWS

Zapier

Apify

DistilBert

ViT

Classwise undersampling

Weighted loss strategies

SHAP

LIME

GitHub

Languages

Urdu

English

Punjabi

Projects

COGNI-KD - MSc Thesis Research

Researcheracademic

Designed multimodal teacher-student deep-learning models for cognitive-overload detection from EEG and physiological signals. Investigated cross-domain knowledge distillation to improve generalization across heterogeneous datasets. Compressed a >1M-parameter teacher into a ~38K-parameter student for wearable deployment while matching or exceeding teacher accuracy. Ran experimental evaluation of architectures and knowledge-transfer strategies in PyTorch, focused on multimodal representation learning and model compression.

PyTorchEEGPhysiological signalsKnowledge DistillationModel Compression

Fake Review Detection & Summarization

personal

NLP pipeline for fraud detection and automated summarization of suspicious reviews (BERT, DistilBERT, T5).

BERTDistilBERTT5NLP

Mindoro — AI-Assisted Productivity System

personal

deep-work system with task orchestration and session tracking, optimizing performance under high cognitive load; ties to overload-detection research (React, TypeScript, Firebase).

ReactTypeScriptFirebase

Your Money Mentor

personal

full-stack AI personal-finance assistant with goal tracking and intelligent insights (Lovable, Supabase).

LovableSupabase

Multimodal Game Genre Classification

personal

BERT/RoBERTa with ResNet/ViT for multilabel genre prediction from text and image data (PyTorch, Hugging Face).

BERTRoBERTaResNetViTPyTorchHugging FaceDistilBert

Healthcare Cost Prediction Semester Project

personal

Explored SHAP and LIME to improve interpretability for clinical stakeholders and increase trust in predictions for a US healthcare model involving patients, providers, and insurance layers.

SHAPLIME

Personal Portfolio

personal

A technical portfolio website hosted at https://fatima-sohail.netlify.app/ showcasing professional work and projects.

Education

MSc Data Science

FAST NUCES, LahoreData Science2024 – 2026

Thesis. Cross-Domain Knowledge Distillation for Wearable Cognitive Overload Detection using EEG and Physiological Signals

BSc Computer Engineering

UET, LahoreComputer EngineeringGPA 3.42020 – 2024

Recognition

Patents & publications

Comparative Analysis of Search Algorithms in AI (2022)

publication10.13140/RG.2.2.29282.611232022

Published.

Read it

Cross-Domain Knowledge Distillation for Wearable Cognitive Overload Detection using EEG and Physiological Signals

publication

New Generation Computing (Springer). Under review.

Hybrid CNN-LSTM for Real-Time ICU Anomaly Detection

publication

Under review

Deep Learning Architectures for Context-Aware Human Activity Recognition in Healthcare Recommender Systems

publication

In preparation.

Hire FatimaReach out about a role, a contract or a conversation.For recruiters

Fatima's twin is AI, it can make mistakes.

More people on Kanvis

Browse more people Create your own Kanvis