Afirah Zahid

Data Scientist with 2+ years of experience spanning GenAI, predictive ML, computer vision, and enterprise AI strategy. At Addo AI, I have built RAG-powered chatbots on AWS Bedrock, developed end-to-end forecasting pipelines on Databricks, and worked on AI maturity assessments for large industrial organizations. Hands-on across the full stack, from YOLO-based vision systems and CatBoost models to LLM integration and organizational AI transformation.

Associate Data Scientist - Addo AI

(2022-10 - 2023-10)

• Conducted AI maturity assessment for one of Pakistan's largest industrial conglomerates, conducting discovery sessions across IT, Finance, Supply Chain, HR, and Marketing to map capability gaps across 6 business functions.
• Delivered a six-pillar AI Current State & Gap Assessment, covering Strategy, Data Platform, Analytics, Predictive ML, GenAI/Agentic AI, and Operating Model.
• Co-designed a technical feasibility scoring framework applied across 260+ AI use cases; prioritization output identified 116 Quick Wins and 84 Strategic Bets to guide the client's AI investment roadmap.
• Built an end-to-end Databricks pipeline (Data Loading → Feature Engineering → Modeling) to predict non-promo baseline sales per mall–product combination, enabling accurate promo lift measurement.
• Engineered time-series features (lags, rolling averages, EWMA, holiday indicators) and trained CatBoost regression models per combination; added a moving-average fallback for low-data products.
• Loaded and joined reference tables in Spark, saved curated datasets to output tables, and documented the full workflow for handover and maintenance.
• Built a POC RAG-powered chatbot using AWS Bedrock Knowledge Base and Claude (Haiku), enabling product Q&A for Sensodyne's product catalog.
• Prepared structured product data (TXT + metadata JSON) for Bedrock KB ingestion; integrated AmazonKnowledgeBasesRetriever with hybrid retrieval tuning for improved search relevance.
• Added price-based and attribute-based filtering to the chatbot flow.
• Built a computer vision-based solution for monitoring office space productivity based on SOTA detection and tracking algorithm. This included creating KPIs such as footfall count, seat occupancy, and heatmap generation of different spaces. Created training and inference pipelines to enhance workflow efficiency. The models were deployed using Flask to create APIs, enabling easy integration with other applications.
• Designed & developed dashboards using Streamlit & Plotly Dash for batch and real-time data monitoring and analysis of the workspace reducing manual reporting time.
• Conducted research and development of new tools and techniques for potential use cases, such as demand forecasting and an attendance system using face recognition.

Associate Data Scientist - Addo AI

(2025-05)

Graduate Research Assistant - RICE Lab, LUMS

(2022-02 - 2022-06)

• Research and development of weed detection and segmentation model using a Mask R-CNN with a ResNet-50 backbone, achieving AP50 scores of 90.47% for cotton and 70.83% for rice. Addressed challenges in distinguishing young weeds from seedlings & variations in crop growth stages. Conducted high-resolution data collection & annotated data using VGG Image Annotator, achieving a generalized model applicable to both crops.

Machine Learning Intern - Department of Computer Science, UET Lahore

(2019-06 - 2019-08)

• Research and development of a wireless module on Raspberry pi3b+ with its powerhouse, screen, microphone, & small speaker achieving 87% accuracy. Used Google APIs and Python libraries for Speech Recognition and Synthesis. Implementation and integration of Speech to Text (STT), Machine Translation (MT), and Text to Speech (TTS) for different languages such as English, Urdu, and Chinese on Raspberry pi3b+.

Master of Science (MS) - Computer Science - Lahore University of Management Sciences (LUMS) (2023-06)

Bachelor of Science (BS) - Computer Engineering - University of Engineering and Technology Lahore (UET) (2020-07)