Conversational Commerce Platform - Hongkong
AI Developer, Jan 2023 – Present
Project description
- Built an enterprise-grade multi-tenant chatbot platform that powers personalized shopping conversations across retail websites using OpenAI's GPT-4 APIs.
Responsibilities
- Designed and built a modular chatbot framework using LangChain and GPT-4, allowing scalable deployment across multiple clients with different intents and personalities.
- Integrated a RAG system with vector search (Pinecone) to enrich chatbot responses with up-to-date product and inventory information.
- Developed a tool-usage orchestration layer enabling the LLM to query APIs like order tracking, recommendations, and cart actions.
- Conducted extensive A/B testing and prompt optimization to reduce hallucination rate and improve CSAT scores.
- Collaborated closely with product teams to iterate on conversational UX and business use cases (e.g., upsell, retargeting).
- Implemented observability pipelines with Prometheus/Grafana to monitor latency, error rate, and fallback performance.
Technologies
- Python, OpenAI API, LangChain, FastAPI, Pinecone, AWS.
Healthcare Virtual Assistant - US
AI Developer, March 2022 – Aug 2024
Project description
- Developed an AI assistant to help chronic disease patients track medication, book appointments, and understand symptoms using domain-tuned LLMs.
Responsibilities
- Fine-tuned and evaluated domain-specific LLMs (BioBERT, MedPalm) to answer patient queries with clinical-level accuracy.
- Designed multi-turn dialog management logic to ensure conversation continuity while managing edge cases like unclear symptoms or conflicting medication.
- Integrated OpenAI Function Calling with Azure-hosted services for medication reminders, appointment scheduling, and doctor Q&A.
- Built a feedback learning loop to retrain the assistant based on real-world interaction logs, improving accuracy over time.
- Ensured HIPAA compliance by implementing strict data handling, masking, and logging policies in coordination with DevSecOps
Technologies
- Python, Pytorch, Triton, AWS, Fast API, Flask, PostgreSQL, Docker, Kubernetes, Redis.
ML Platform for Smart Recruiting - US
AI Developer, Jan 2021 – Jun 2021
Project description
- Created intelligent recommendation engines for a recruiting startup, matching candidates with jobs based on skill semantics.
Responsibilities
- Built and deployed a semantic matching engine between resumes and job descriptions using cosine similarity and FastText embeddings.
- Created an automated skill extraction tool using SpaCy + RegEx with feedback loops from recruiters to improve precision.
- Trained ranking models using XGBoost and LightGBM, optimizing for click-through rate and job acceptance probability.
- Developed a web-based internal analytics dashboard to visualize matching insights and user drop-off trends.
- Collaborated with frontend engineers to implement an interactive job suggestion assistant for candidates.
Technologies
- Python, Git, Transformers, BERT, Flask
AI Chatbot for Banking Support - Vietnam
AI Developer, May 2018 – Oct 2020
Project description
- Developed and deployed a Vietnamese-English bilingual chatbot for one of the top digital banks in Southeast Asia.
Responsibilities
- Engineered an intent classification and entity extraction system using a fine-tuned BERT model for 90+ banking use cases.
- Integrated chatbot with backend APIs to automate tasks such as balance inquiry, loan eligibility, and transaction alerts.
- Created a fallback mechanism for ambiguous inputs, using decision trees and heuristics to recover gracefully from failed intents.
- Localized the bot for Vietnamese language using VN-BERT and rule-based sentence normalization for dialect support.
- Deployed and maintained the solution in production using a microservices-based architecture on Kubernetes.
Technologies
- Python, BERT, Rasa, Docker, Kubernetes, Firebase
