- Orchestrated the design and implementation of a Document-Based Question Answering system leveraging LangChain, Open AI LLM, and Pinecone to alleviate the workload for support teams with MEVN stack.
- Experience on create, modify, and retrieve calendar events with Google Calendar API to create meeting, using 100.ms library to get script text of the conversation.
- Led the design and development of Promp Engineering:Summarization, Classification, Translation, Text Generation, Question &Answering, Coaching … to control output :length, style, audience, context, scenario
base guiding, chain of thought, avoid hallucination .
- Do research on Llama 2, fine tune Last Layer, Adapter Layers, Prefix Tuning, LoRA Low Rank matrix train (less parameters, training time, RAM usage, switch between different LoRA weights), QLoRa (4 bit optimal)
for privacy purpose and improve performance.
- Build chat application with Open-Source LLM MPT-7B locally to protect privacy information for customers better answer and lower response time, reduce cost.
- Successfully refactored complex code, reducing its complexity and enhancing readability : Refactor code from 5 embedded dimensions of if-else condition to 1 dimension, implement local processing instead of
posting to API server itself. Writing different functions in utils and services folder which can be re used in the application. It follows SOLID, DRY, KISS priciples . Other developer failed to do this.
- Processing data for faster retrieval by using local storage, caching with redis, store file on S3, store data to database and creating indexes for quick query…
- Writing background task to synchronize email data with Database, S3 for quick retrivement. Writting producer and consumer to exchange message between micro services.
- Using Helm chart to dynamic inject variable for Kubernetes deployment.
- Using ORM Alchemy to make code clean, readible, easy to switch to other DB such as MySQL, Postgres…
- Organize source code in services, models, utils, controllers which can be used across the applications.
- Implemented testing, monitoring, debugging, reading code, reviewing code, communicating with other developers on design, guiding interns on coding, writing documents,
Technologies: Python, Django, Vue.js, TypeScript, AWS, Lambda, EC2, S3, CloudFormation, ECS, EKS, SQS, SNS, DynamoDB, Docker, Terraform, Kubernetes, Travis CI, GraphQL, PostgreSQL
