Fahad demonstrates expertise as a seasoned back-end engineer, boasting a remarkable track record of over two years in crafting high-performing REST APIs, implementing robust queueing architectures, and deploying efficient caching protocols across diverse applications. He possesses extensive proficiency in Python and holds commendable command over leading frameworks such as Flask and FastAPI, adeptly navigating through eminent technologies including SQLAlchemy, PostgreSQL, MongoDB, RabbitMQ, and Redis. Additionally, Fahad exhibits invaluable versatility by incorporating DevOps practices with Docker, AWS, and Azure. Through his meticulous approach, he ensures the development of modular and extensible solutions, all while effortlessly aligning with functional and business requirements. His unwavering commitment to delivering exceptional products is consistently demonstrated by their potential to revolutionize business processes and drive enhanced profitability.
Led the development of an MVP marketplace application for the Black community.
Architected backend infrastructure using FastAPI, Python, and Elasticsearch, and integrate Stripe for payment and subscriptions. Developed the frontend using Next.js.
Built large scale jobs for entire website crawling, using rotating proxies, selenium, mocked chromium browsers, and added extensive NLP pipelines (tokenization, embeddings) to feed into Elasticsearch indexes.
Motive
ML Operations Engineer
2024 - Present (1 year)
Remote
Engineered and implemented high-performance, scalable inference workloads utilizing Kubernetes and NVIDIA Triton, achieving a 15x reduction in AI inference costs and a 10x improvement in processing speed. Enhanced service visibility through the integration of Grafana, Prometheus, and Datadog.
Revamped Motive's Customer Visualization pipelines in C++, leveraging OpenCV, FFmpeg filters, and Substation Alpha subtitles format to decrease video processing and encoding times by 5x.
Created Model Training and Testing workflows using Kubeflow and Anyscale (Ray), incorporating model and experiment tracking functionalities with MLFlow.
Currently focused on implementing agentic workflows utilizing LLMs, Embedding Models, and Vector Stores (Qdrant) to facilitate video retrieval, event reporting, and various other operations.
The Entertainer FZ
Senior Software Engineer
2022 - 2024 (2 years)
Remote
Developed RESTful microservices using Flask, SQLAlchemy, MariaDB, and Elasticsearch with Session and JWT Authentication for various product features.
Engineered a dynamic multi-tenant database connection in APIs to reduce code redundancy, DB overheads, thus enhancing performance by 30%.
Built Data Warehouses to consolidate data from Salesforce, SQL/NoSQL databases, S3 storage, Clickhouse DBs into Snowflake and Elasticsearch indices.
Led migration of all microservices to Python 3.11 and Poetry for better dependencies control, along with upgrading obsolete packages and removing deprecated code.
CureMD
Artificial Intelligence Engineer
2020 - 2022 (2 years)
Remote
Engineered and developed an automated Resume Parsing and Scoring system using TensorFlow, PyTorch, PaddlePaddle, Tesseract, and Spacy. Developed multiple AI and ML pipelines to understand candidate profile and highlight prospective candidates.
Designed a multi-threaded RabbitMQ-based audio transcription and hold-detection system, achieving 85% ASR and Hold Detection accuracy using NVIDIA Nemo ASR and multiple NLP pipelines.
Led the development of an automated Fax OCR system for patient info and lab reports, reducing data entry errors and digitizing medical records, using RabbitMQ and Celery.
Developer created an automated hiring platform that significantly reduces lead time. Used Flask and Docker for backend development, deployed on AWS ECR and Fargate, integrated with a load balancer for increased availability, and MongoDB for the database.
Constructed a system pipeline involving document parsing, data normalization, candidate grading and deployment. Built a reliable PDF parser using MuPDF to extract important features from documents and developed a prediction model for data segmentation based on these features.
Trained a spaCy NER model (BERT) on education and experience data to extract key details such as organization, dates, and roles. Designed textual libraries using the FlashText algorithm to swiftly extract over 50,000 skills, normalize 3,000+ institutions, and provide accurate location data.
Designed and developed an information extraction system to automate patient registration workflows by extracting personal data from insurance cards. Reduced time and errors associated with manual data entry.
The system was deployed as an Azure Function and integrated with an Android application. Utilized OpenCV for card region identification from raw images, improving text prominence and accuracy.
Trained optical character recognition (OCR) pipeline using PaddleOCR, resulting in a significant increase in text detection accuracy, from 55% with Tesseract to 93%.
Successfully developed an AI-assisted dialer bot that efficiently handles operator queries, transcription tasks, and recognizes hold and pick-up actions, significantly enhancing efficiency.
Analyzed various open-source automatic speech recognition (ASR) frameworks including Vosk, SPINE, OpenVINO, and NeMo, and evaluated their Word Error Rate (WER) performance on internal insurance voice calls.
Devised and executed a messaging-based queuing protocol system using RabbitMQ, that allows the system to manage up to 40 concurrent calls for transcriptions, identifying hold durations, call cancellations, and handling of interactive voice responses (IVR).
Developed a platform to assist black-owned businesses in registration, validation, and integration with an expanding search engine, including creating administration, business, and user portals
Implemented role management and permission protocols, and constructed the application using technologies like React, Python FastAPI, and storage backends such as MariaDB, MongoDB, and Elasticsearch
Created a comprehensive scraping pipeline for data gathering from diverse sources, like Instagram, utilizing JavaScript and Python libraries for optimized user experience
