AI Engineer building reliable GenAI products with a product-first lens.
I combine 4+ years of software delivery experience with hands-on AI engineering and product strategy. My work focuses on model evaluation, LLM reliability, latency, cost, user trust, and translating technical systems into polished product experiences.
Featured work
This portfolio highlights both AI research and human-centered engineering: one project evaluates model quality and deployment tradeoffs, while another combines sensing, 3D prototyping, and ML classification.
NLP research / LoRA / evaluation
Co-authored research on educational text difficulty classification. We reproduced a 63-prompt readability baseline and compared faster alternatives, including scalar-mixed ELECTRA, DANN, and LoRA-tuned Phi-3.5-mini.
Hardware + ML / Human factors
Built a passive wrist-support prototype using 3D-printed components, lab sensor data, and ML classification. The project connects human-centered design with physical prototyping and model-driven analysis.
AI systems
The project set shows hands-on engineering, product discovery, go-to-market thinking, recommendation systems, and human-centered ML work.
GenAI product / RAG
An agentic career platform with job matching, portfolio generation, outreach drafting, skill simulation, RAG Q&A, and agent chat. Built with React, Express, FastAPI, ChromaDB, and NVIDIA NIM.
AI product discovery
An AI developer reporting concept shaped through 25+ interviews with developers, engineering leads, and managers. I translated discovery insights into positioning, prioritization, pricing, and prototype storytelling.
Product strategy / GTM
A smart commercial lighting concept focused on customer personas, KPI definition, market sizing, pricing, and distribution strategy. The project connects customer value, business viability, and go-to-market execution.
ML learning system
A speech-to-video learning pipeline for accented-English learners. It combines speech recognition, keyword extraction, and recommendation logic to make educational content easier to find.
Interactive learning resource
A structured study resource with interactive notes, formula references, and lecture-based summaries designed for revision, concept review, and exam preparation.
Hardware + ML / Human factors
A passive wrist-support system combining 3D prototyping, sensor-based lab studies, and ML classification to reduce physical strain. The work connects technical experimentation with human-centered problem solving.
Product lens
Strong AI products come from disciplined product judgment: clear problem framing, meaningful evaluation, user trust, workflow fit, and deployment economics.
Problem framing
Understand the user decision, current workaround, and cost of failure before choosing a model or designing the feature surface.
Evaluation
Accuracy is not always enough. For AI products, robustness, trust, latency, cost, and out-of-distribution behavior often matter more.
Trust
Users need to know when an AI output is reliable. Good AI UX includes citations, uncertainty, guardrails, review flows, and recovery paths.
Deployment
Model quality must be balanced with inference time, energy use, integration effort, data quality, and the user's tolerance for delay.
Experience
Recent work combines applied LLM product engineering with the software delivery foundation needed to ship reliable systems.
Current internship
AI Engineer Intern
Engineering foundation
Software Engineering and Technical Delivery
Shipped software, automated workflows, and coordinated technical delivery across distributed teams. This experience shaped the feasibility and operational judgment I bring to AI product work.