AI Art Studio
🏆 1st Place — Harvard SEAS LLM Hackathon
Award Winning AI Art Studio converts a single prompt into a short cinematic clip.
Generated images and video, a narrated script, and synchronized sound.
It opens new worlds for content creators, storytellers, educators,
and filmmakers, who can prototype scenes and emotions in seconds.
The system orchestrates four task specific models (visual, narrative,
audio, and a QA auditor), applies accelerated performance,
methods grounded in 70+ of AI research.
Vizor – Pharmacy Verification Computer Vision for Robotics
Vizor is an enterprise grade computer vision and sensor fusion platform for automated pharmacy verification in robotics and medical technology.
The system uses a ConvNeXt Deep learning model together with sensor data to verify medication trays, labels, and doses, achieving accuracy with over 99% accuracy
and latency under 100 milliseconds on GPU and Apple MPS.
Vizor includes a FastAPI inference service, robust safety logic with fail closed behavior,
a reproducible evaluation pipeline, and a design that is ready for production with a complete model card and dataset management.
Sophia AI - A Standalone Local LLM for Technical Interview Coaching (100% Offline, No Cloud Calls)
Sophia AI is a domain specific, fine tuned Large Language Model designed to act as an expert technical interviewer
for AI and ML, backend, and MLOps roles. Unlike generic wrappers, Sophia follows a full ownership approach:
she is not an API call to OpenAI, she is a standalone 1.5B parameter neural network fine tuned with QLoRA on a
proprietary Enterprise Knowledge Lake, with synthetic alignment to reduce common hallucinations. It supports local
inference on Apple Silicon or NVIDIA GPUs, ensuring zero external API dependencies.
MiahAI – Offline, Full-Stack AI Platform for Private Learning
MiahAI isn’t just another chatbot it’s a production ready AI learning application to teach,
not just hand out quick answers or behave like most chatbots do like a calculator.
Instead of giving you the answer instantly, MiahAI provides context grounded in proven
teaching methods shaped by decades of research on how people truly learn and retain knowledge,
so the user can actually learn.
Offline Modular Reasoning Engine: Testing Miah LLM
Client agnostic LLM reasoning service with server streaming gRPC,
hot swappable model adapters (OpenAI, DeepSeek, Grok, Azure APIs and a local model of AI Miah(LLM)
via ONNX Runtime and TensorRT for LLMs), and SQLite session memory.
Ships with health and readiness endpoints, structured logging, unit tests, and a Python client.
Experience
About 4 years of hands on experience working with software and coding: 2 recent years focused on Artificial Intelligence
and Full Stack Software Engineering, plus 2 earlier years in high school programming classes.
Education
-
Harvard University — B.L.A., Extension Studies (CS)
Harvard Extension School
Expected 2027
-
Harvard SEAS — Large Language Models: From Transformer Basics to Agentic AI (Professional Program)
Harvard John A. Paulson School of Engineering and Applied Sciences
Oct 2025
-
HarvardX — Machine Learning and AI with Python (in progress)
HarvardX
2025
-
HarvardX — Professional Certificate in Computer Science for Artificial Intelligence (Grade A)
HarvardX
2025
-
TrueCoders — Full-Stack Software Engineering Certificate
TrueCoders
2025
-
Cuesta College — General Education(Math)
Cuesta College
2013–2015
Professional Experience
-
Machine Learning Engineer — AI Art Studio (1st Place)
Harvard SEAS LLM Hackathon, project collected by Meta
Oct 2025
Led a team of four to build AI Art Studio, a prompt to cinematic video pipeline that turns a single input
into visuals, a narrated script, and synchronized audio with support for re editing. Trained and integrated
a “cinematic director” quality assurance model in Python using scikit learn, TensorFlow, and PyTorch to
orchestrate multiple models and enforce consistency, and shipped a live demo in a three hour build window.
-
AI and Machine Learning Engineer
Bright Minds Private AI
2024–Present
Design and build standalone large language models and private AI systems that run fully offline and on premises.
Own the full path from data preparation and fine tuning through evaluation and deployment, including modular
reasoning engines and high performance inference services built with C sharp, Python, gRPC, and relational databases.
Use Microsoft Azure and Amazon Web Services for experimentation, data workflows, and cloud baselines while keeping
the primary systems designed to operate without external dependencies. Focus on clean architecture, observability,
and deployment patterns that are ready for enterprise environments.
-
Software Engineer | Applied AI
TrueCoders — Full Stack Software Engineering
Sept 2024–Nov 2025
Built MiahAI, an offline private learning assistant: an ASP.NET Core MVC chat application with persistent
session memory, CRUD workflows, and a streaming style interface. Implemented layered architecture across
web, core, and infrastructure projects with dependency injection, validation, and exception handling,
backed by Entity Framework Core and MySQL migrations. Integrated a typed gRPC and Protobuf boundary to a
local modular reasoning engine for low latency model calls and stable contracts.
