FlowCV Logo
Letlhogonolo Fanampe AI/ML Engineer | Computer Vision | Gen AI | NLP
Profile

AI/ML Engineer with a deep passion for developing innovative solutions using Generative AI and NLP technologies. Experienced in building, fine-tuning, and deploying LLMs for diverse applications, including fake news detection, creative text generation, and multimodal interactions. Proven track record in winning hackathons and delivering high-impact AI solutions. Certified in Python development, Generative AI, and web development, with a strong foundation in analytical chemistry.

Professional Experience
AI/ML Testing Engineer - Generative AI & NLP, Next Sapien
January 2024 – June 2024
  • Role Summary: Developed and implemented rigorous test protocols for AI/ML projects focusing on Generative AI, NLP, and computer vision. Ensured high accuracy rates of 96% in development and 99% in production environments.
  • Key Contributions:

  • Mobile App Input Field Characterization: Extracted and validated input field data from mobile apps using ADB, ensuring accuracy against the expected JSON schema.
  • Android App Navigation Chatbot: Tested AI chatbot navigation within Android apps using natural language commands, focusing on accuracy and efficiency.
  • Message Extraction: Enhanced text, timestamp, and emoji extraction from chat screenshots with 99% accuracy, ensuring proper data categorization in JSON format.
  • Human Characterization API: Tested physical attribute estimation and subjective trait inference from human images, addressing ethical considerations.
  • Code Summarization & Querying: Evaluated code summaries for Angular projects and integrated them into a custom LLM model for accurate project structure queries.
  • Multimodal Data-Integrated Chatbot: Assessed chatbot performance in retrieving and integrating data from Graph, JSON, and NoSQL MongoDB APIs.
  • Developed and enhanced facial recognition algorithms, significantly improving the accuracy and efficiency of computer vision applications.
  • Pioneered the integration of height estimation in facial analysis, achieving a Mean Absolute Error of 6.6 cm, setting new standards for precision in height prediction algorithms.
  • Led efforts in optimizing Optical Character Recognition (OCR) technologies, enhancing text extraction accuracy to over 95%, thereby improving the efficacy of chat analysis tools.
  • Collaborated with interdisciplinary teams to implement and refine AI/ML solutions, contributing to the development of innovative applications and functionalities.
  • Maintained cutting-edge knowledge in the field of AI/ML and computer vision, regularly updating team methodologies and contributing to a culture of continuous learning and improvement.
  • January 2023 – present

    - Developed AI-driven fake news detection system using LLaMA 3 model, achieving 95% accuracy in classifying real-time news articles.

    - Built and deployed Generative AI-based Telegram bot leveraging Google’s Gemini-pro LLM API, enabling creative text generation for over 1,000 users.

    - Created a multimodal interaction bot using Google’s Gemini Flash model to process and respond to both text and images, enhancing user engagement in a Streamlit app.

    - Fine-tuned and optimized LLMs for diverse NLP tasks including sentiment analysis, text summarization, and creative writing.

    Education
    National Diploma: Analytical Chemistry, Tshwane University of Technology
    January 2010 – July 2015Pretoria, South Africa

    Completed coursework in analytical chemistry, inorganic and organic chemistry, physics, and mathematics. Developed hands-on laboratory skills in practical chemistry courses. Demonstrated proficiency in chemical quality assurance and computer skills.

    National Senior Certificate: Grade 12, Pretoria Technical High School
    January 2005 – December 2009Pretoria, South Africa

    Completed secondary school education with coursework in English, Afrikaans, mathematics, life orientation, electrical technology, information technology, and physical sciences.

    Languages
    English

    Fluent proficiency.

    Setswana

    Native proficiency

    Skills
    Programming Languages

    Python, JavaScript, HTML, CSS

    Generative AI/NLP

    LLMs, LangChain, OpenAI API, Whisper, Google's Gemini, Meta’s LLaMA, Ollama

    Data Management

    SQL, PostgreSQL, Pandas, NumPy

    AI/ML Frameworks

    TensorFlow, PyTorch, Keras, Hugging Face Transformers

    Tools and Technologies

    Streamlit, Flask, FastAPI, Git, Docker

    Certificates
    LangChain & Vector Databases in Production

    For my achievements in Gen Al 360 Certification, developed in collaboration with Activeloop, TowardsAl, & Intel Disruptor Initiative.

    WebGPU Hackathon - Outstanding Performance by LabLab.ai

    Recognized for outstanding performance, attendance, and successful completion & submission of a WebGPU solution.

    Data Visualization Developer Certification

    Demonstrated proficiency in various data visualization techniques, creating interactive visual representations of data to communicate insights effectively.

    Mimo Web Development & Programming Mastery

    Acquired proficiency across multiple domains including HTML, JavaScript, Python, SQL, and end-to-end web development. Demonstrates a holistic understanding of web creation, programming languages, and database management.

    Autonomous Agents Hackathon

    For outstanding performance, attendance, successfully completing & submitting a solution based on LangChain, BabyAGI, Auto-GPT, Weavite, Qdeant, SuperAGI

    Career Essentials in System Administration by Microsoft and LinkedIn

    Learnt the essentials of system administration.

    Responsive Web Design Developer Certification

    Acquired skills in designing responsive web layouts, ensuring web content is optimally viewed across various devices.

    MS Excel Fundamentals & Intermediate 2010

    Mastery of foundational and intermediate Excel functions and features, including data visualization using Excel charts and graphs.

    WebGPU Hackathon - Best Solution by LabLab.ai & New Native Group

    Awarded for delivering the best solution during the WebGPU Hackathon.

    JavaScript Algorithms and Data Structures Developer Certification

    Successfully completed the JavaScript Algorithms and Data Structures Developer Certification. Demonstrated proficiency in algorithmic challenges and data structure projects.

    Front End Development Libraries Developer Certification

    Demonstrated proficiency in various front-end libraries and tools through hands-on projects.

    LECO SC632

    Front-Line maintenance and operational training course

    Courses

    Comprehensive training on foundational AI model theory, with hands-on experience in LangChain and Deep Lake technologies, emphasizing the application of Large Language Models across industries.

    Master Python 3, Udemy
    August 2023

    This comprehensive course is designed for individuals ranging from complete beginners to advanced programmers who wish to master Python 3. The curriculum delves deep into Python, starting with an introduction to the language, its documentation, and transitioning from Python 2 to Python 3. Participants will be guided through the setup process, including the installation of Python and various IDEs. The course then covers Python basics such as syntax, data types, loops, functions, and more. Advanced topics include working with different data structures like strings, lists, tuples, sets, and dictionaries. By the end of the course, students will have a thorough understanding of Python 3 and its applications.

    This course provides an in-depth understanding of system administration, covering essential skills and tools required for the role. It offers insights into managing various implementations, including on-premises, cloud, physical, and virtual, as well as storage security. The course is designed to equip learners with the knowledge to advance their career in system administration and earn a Career Essentials Certificate by Microsoft and LinkedIn.

    This course offers an in-depth exploration of the Unified Modeling Language (UML) and object-oriented design principles. It covers the fundamentals of software development methodologies, with a focus on Waterfall vs. Agile, and dives deep into object-orientation concepts such as Abstraction and Polymorphism. The course provides hands-on experience in mapping requirements to technical descriptions, understanding UML diagrams, and designing a Note-Taking App from scratch.

    This course delves into the Bubble ecosystem, a leading No-Code platform, teaching how to create professional web applications without coding. Topics covered include responsive design, database management, workflows, user management, and advanced features like plugins. Hands-on projects included building a project/task management app and a simplified version of AirBnB with integrated features.

    I completed the "Build Your Startup with No Coding (Design, Develop & Ship)" course on Udemy. Led by an experienced Bubble developer and coach, the course provided hands-on experience in building a project and task management application named 'Teambase'. The curriculum covered a wide range of technical and design skills, including understanding the Bubble Editor, page layout, design, workflow logic, database design, and live app deployment. By the end of the course, I transitioned from a beginner to possessing an intermediate skill level in Bubble development.

    This course offers a comprehensive guide on how to create a ride-sharing app similar to Uber using Bubble's no-code platform. It covers the entire development process, from database configuration, user registration, integrating Google Maps API, to processing payments and user reviews. The course also provides insights into Bubble's toolset, creating custom databases, and user account management. Ideal for aspiring entrepreneurs, no-code enthusiasts, and anyone interested in building a ride-sharing platform without coding.

    This course covered Bubble technology, enabling the creation of applications without code. Topics included responsive page creation, database design, CRUD operations, workflow logic, custom states, dynamic data integration, and more. It aimed to equip learners with the skills to develop applications without the need for traditional coding.

    This course is led by an experienced Bubble developer and coach who has a track record of assisting founders in creating successful applications using Bubble's no-code technology. The course narrative is rooted in the instructor's personal journey of discovering Bubble in 2016 and leveraging it to launch a startup within a year. Participants will gain hands-on experience in building a project and task management application named 'Teambase'. The curriculum covers a wide range of technical and design skills, from understanding the Bubble Editor, page layout, and design, to mastering workflow logic, database design, and deploying the app live. By the end of the course, learners will have transitioned from beginners to possessing an intermediate skill level in Bubble development.

    This comprehensive course on Bubble's visual programming platform covered building web and mobile apps without coding. Topics spanned from user interface and experience design, payment gateways, databases, workflow logic, API integrations, to deploying apps. The course also included hands-on projects, such as building a Tinder clone app, and provided templates for various applications.

    Projects

    This project is a Python-based script designed to accurately extract text, timestamps, and emojis from chat conversation screenshots. The primary goal of this project is to achieve a character-level accuracy of 99% in text extraction, ensuring reliable and precise results.

    After rigorous testing and evaluation, the project achieved the following results:

  • Character Error Rate (CER): 0.97% (99.03% accuracy)
  • Word Error Rate (WER): 1.12% (98.88% accuracy)
  • Timestamp Detection Accuracy: 99.2%
  • Emoji Detection Accuracy: 98.7%
  • An innovative AI project for NextSapien, focusing on advanced facial analysis. This repository contains scripts and methodologies for gender identification, face rating, and comprehensive demographic insights using AI and computer vision techniques. The project showcases a unique blend of analytical rigour and creative problem-solving, aligning with NextSapien's ethos of pushing the boundaries in AI/ML applications.

  • Developed an application that suggests Disney songs based on user-entered emotions.
  • Utilized DeepLake, LangChain, and Upstash for data storage and retrieval.
  • Implemented emotional embedding and similarity search algorithms.
  • Primary language: Python.
  • Finetuned the Llama 3 70 B model using Together AI
  • Fined tuned on 120 000 Dataset
  • Developed an AI-driven system using Together AI's fine-tuned LLaMA 3 model.
  • Real-time classification of news articles as real or fake with 95% accuracy.
  • References
    Robin (Bintang) T, Regional PM at Sea | ex. Astra | Lablab.ai Hackathon Team Member
    Mudassir Fayaz, Mentor | Hackathon Team Leader | Frontend Engineer | Hackathon Winner | BSCS'24 | AI & ML Enthusiast | medium writer | leetcoder, https://www.linkedin.com/in/mudassir-fayaz-8a7697204/
    Anant Bedi, Senior Developer, Next Sapien
    Gift Olifant, Lab manager, Afrisam