Solidity Developer

As the lead developer at Kredora, I was responsible for creating a platform that conducted credit scoring for microloan applicants. This involved integrating third-party services to perform background checks, KYC processes, and validate the authenticity of identity documents, income statements, and address proofs. I use a fine tuned Llama 3 model, to extract data fields from income statements, address proof documents and government IDs. I implemented AuraFace model to compare the similarity between an applicant's ID photo and a video selfie, ensuring accurate identity verification.

Designed and deployed an NFT Marketplace smart contract on the Avalanche network using Solidity, OpenZeppelin libraries, and Hardhat. This allowed users to trade NFTs using ERC20 tokens and AVAX. Developed multiple subgraphs to index transactions, trends, and analytics for all ERC721 contracts on the Avalanche network, providing users with actionable insights for navigating NFT collections. Automated CI/CD processes using GitHub workflows, ensuring efficient testing, compilation, and deployment of new Web App features. Created a microservice to index NFT metadata for all collections on the Avalanche network, storing it in AWS S3 and serving it through CloudFront to achieve low-latency performance compared to IPFS. Led backend development using a microservices architecture with Node.js and Docker.

Enhanced Conversational AI with ML Models: Spearheaded the implementation of machine learning models to discern patient intentions during conversations, significantly boosting LENA's conversational AI capabilities. Applied Transfer Learning for NLP Models: Employed advanced transfer learning techniques to retrain open-source NLP models, such as GPT-2, using proprietary datasets, optimizing their performance for healthcare applications. Developed and Deployed AI Chatbot MVP: Successfully conceived and launched a minimum viable product (MVP) AI chatbot capable of engaging with patients, adeptly identifying their intentions, prompting relevant follow-up questions, and delivering pertinent information. Created LENA Health Assistant Platform: Pioneered the development of the LENA Health Assistant platform, seamlessly integrating Patient Relationship Management (PRM) and Case Management tools to provide an all-encompassing solution for personalized patient care.

Led the development of the Andia SDK's mobile framework for iOS and Android platforms, enabling seamless integration of digital KYC onboarding, using biometric functionalities. Develop a Machine Learning model based on architectures like ResNet and DenseNet, to create highly accurate classifiers for biometric identification (Face Recognition). Implemented a Convolutional Neural Network (CNN) using TensorFlow and Keras, establishing a robust biometric model for liveness detection, to avoid biometric spoofing (fraud). Developed RESTful APIs consumed by mobile clients, ensuring high availability and fault tolerance through efficient deployment strategies. Andia was accepted to be part of Techstars WU 2019 cohort in Denver.

Designed and implemented text processing models capable of extracting relevant information from massive volumes of text, enhancing data insights and analysis. Created a scalable backend system to handle a large number of images and facilitate text processing tasks. Developed and managed RESTful APIs using Python and Tornado, providing seamless access to the deployed machine learning models. Developed machine learning models using Keras and TensorFlow, specializing in sequence-based (text) data analysis and processing.

Created and trained a model for predictive forecasting of foot traffic in physical retail stores, enabling informed decision-making for retailers. Developed a high-performance platform capable of processing millions of records per minute, demonstrating scalability and efficiency. Pioneered the development of one of the first heat map solutions for foot traffic in retail stores using WIFI-derived data, providing valuable insights into customer behavior. Implemented an open-source technology stack platform, leveraging MongoDB, AngularJS, Node.js, RabbitMQ, Redis, and Hadoop, to enable scalable and flexible data processing.

Developed the first earthquake alert app for Mexico City, integrating with the public earthquake alert system to connect millions of smartphones. Designed and implemented a high-availability platform capable of supporting millions of mobile users. Built the backend infrastructure to enable traffic alerts. Developed and launched the app’s initial versions for iOS and Android. Key Achievements: Secured over 2 million active users. Successfully provided proactive alerts for more than six major earthquakes in Mexico City, delivering warnings with over 50 seconds of advance notice.