Karim Wahby

Company Overview: Outsourced to knowis
- First AI Hire for the Company, guiding them through AI and ML adoption.
- Introducing AI Feature stages and aligning them with the companys SDLC
- Educating team members about AI and ML
- Taking Architectural Decisions on how to implement features and the tools to be used.
- Documenting Architecture Decision in the form of ADRs, ensuring both Comprehensiveness of the decision and Knowledge Transfer.
- Helping PO understand the visibility and complexity of features required, aligning for a phased release approach, and deciding priorities using the 20/80 rule. Allowing us to deliver the first feature in under three months.
- Helping the team understand the difference between POC and MVP features and what it takes to drive the feature to a production-ready state with MLOps in place.
- Building an Architecture Decision Record (ADR) Writing Assistant (Aka: ADR Sidekick) that helps the user in the ADR Authoring process and analyzes dependencies and conflicts between the current document and other company technical documents.

• First AI Hire for the Company, guiding them through AI and ML adoption.
• Introducing AI Feature stages and aligning them with the companys SDLC
• Educating team members about AI and ML
• Taking Architectural Decisions on how to implement features and the tools to be used.
• Documenting Architecture Decision in the form of ADRs, ensuring both Comprehensiveness of the decision and Knowledge
Transfer.
• Helping PO understand the visibility and complexity of features required, aligning for a phased release approach, and deciding
priorities using the 20/80 rule. Allowing us to deliver the first feature in under three months.
• Helping the team understand the difference between POC and MVP features and what it takes to drive the feature to a production
ready state with MLOps in place.
• Building an Architecture Decision Record (ADR) Writing Assistant (Aka: ADR Sidekick) that helps the user in the ADR Authoring
process and analyzes dependencies and conflicts between the current document and other company technical documents.

A Summer Intern at Microsoft Advanced Technology Lab in Cairo (ATLC) as part of the LUIS (Language Understanding Intelligent
Service) Team for three months.

My role was to implement the Data Reduction Algorithms, I implemented (from Scratch) both Johnson Lindenstrauss lemma and
ISOMap, as means to reduce data that lies in both Euclidean and non-Euclidean spaces

Led a cross-functional team of 5 engineers across data pipelines, model tuning, backend services, and ad-hoc requests. Delivered features in phased rollouts while aligning leadership expectations and capturing critical market opportunities.
Re-architected systems and team processes to eliminate single points of failure, increase team productivity, and allow for more exposure and ownership for team members.

- Optimizing the Service costs by 6.5X from few thousand dollars a month to few hundred.
- Re-designing the entire Solution, implementing better separation of concerns, stabilizing the services, enabling both horizontal and vertical scaling, and reducing the operational cost.
- Fine Tuning our Product AutoTagging solution, and enabling RLHF.
- Redesigning the Market Scanning Pipelines to be fully automated and monitored. (Main Data Collection Pipelines)
- Eliminating Single Points of Failure from the system architecture and documenting existing services and procedures to eliminate the threat of Institutional Knowledge being lost.
- Building The Following e-commerce and fashion-focused components:
- Neural text search
- Image search
- Product Listing Categorization
- Product Auto Tagging
- Similar item recommender
- Dynamic Pricing
- Pricing Assistant
- Product Listing Appeal
- A set of Market Scanning Pipelines.

Led a cross-functional team of 5 engineers across data pipelines, model tuning, backend services, and ad-hoc requests.
Delivered features in phased rollouts while aligning leadership expectations and capturing critical market opportunities.
Re-architected systems and team processes to eliminate single points of failure, increase team productivity, and allow for more
exposure and ownership for team members.

• Leading a team of 10 AI Engineers. Mentoring the diverse technical team, bringing out the best in them they could be, while
encouraging their differences and focusing on their strength points, ensuring excellence, collaboration, exposure, and
ownership.
• Providing the team with technical direction, hands-on designing, implementing, deploying, testing, fixing, and prioritizing
investments to build the following:
◦ Dialectical Arabic and English TTS with Voice Cloning feature
◦ Knowledge Mining
◦ Language Classification
◦ Intent classification Training service
◦ Entity Extraction Training service
◦ Entity Normalization service

• Guiding a team of Junior Data Scientists.
• Architecting and implementing a home Medical Chatbot.
• Responsible for implementing, evaluating, deploying, maintaining, and improving each of the following:
◦ Intent classification
◦ Entities extraction and normalization
◦ Language classification
◦ Labeling tool and the labeling experience
◦ Statistics for the chatbot performance
◦ ASR (Automatic speech recognition) Service
◦ Salesforce (CRM) integration
◦ Medical Diagnostic and Triage service integration

Responsible for different teams with more than 20 people between ai, product, and dev ops engineers.
Hands-on in the process starting from customer requirements or user stories to solution deployment and bug fixes.

Key Achievements:
- Managing cross-functional teams comprising over 20 AI, product, and DevOps engineers.
- Revamped the system to support offline deployment on Kubernetes clusters and IIS servers.
- Made critical System Architecture changes to implement separation of concerns within our components.
- Delivered our solution to banks, successfully navigating the security clearance process.
- Maintained the roadmap delivery while migrating revamping the solution
- Successfully Migrated all existing clients to the new solution and codebase
- Covered for missing capacity within engineering until successful hires were made.
- Expanded the engineering and DevOps teams to meet market needs.
- Directed the end-to-end process from customer requirements to solution deployment and bug fixes.
- Spearheaded the development and implementation of advanced conversational AI solutions across all teams.
- Enhanced team productivity and project delivery through effective leadership and strategic planning.
- Mentoring the team and guiding them through the weak points.

Projects:
- Guiding a team of Junior Data Scientists.
- Architecting and implementing a home Medical Chatbot.
- Responsible for implementing, evaluating, deploying, maintaining, and improving each of the following:
- Intent classification
- Entities extraction and normalization
- Language classification
- Labeling tool and the labeling experience
- Statistics for the chatbot performance
- ASR (Automatic speech recognition) Service
- Salesforce (CRM) integration
- Medical Diagnostic and Triage service integration

A member of the NLP Team at ATLC
Worked on the following:

Knowledge Extraction from Documents
• PDF Parsing (Graphical component extraction and collated text separator injection)
• Table detection (detect tables within PDF and extract them).
• Table Layout Understanding (number of rows, number of columns, and cell spanning etc).
• Table joining from across pages
• Table content understanding (Identifying Subjects, Objects, and Relations from within the tables and their context).
• Table representation: converting the understood information from the table and the actual table into either questions and
answers, OR segments of the answer

CCI - Virtual Agent Platform for the support domain
• Customer deployment service (deploy Toronto to the customer subscription)
• Access control (Authentication and Authorization) across all portals, also setting up initial users at the customer subscription.
• All User Administration pages.