AI Software Engineer

• Developed a fully concurrent pipeline integrating state-of-the-art Retrieval-Augmented Generation (RAG) techniques with vector databases, boosting retrieval speed by 80%.
• Designed and deployed an end-to-end production-ready API with FastAPI, ensuring scalability and seamless integration of chatbot solutions.
• Built multilingual data collection pipelines for fine-tuning embedding models, reducing warm-up time to <1s on a single A100 GPU.
• Fine-tuned and pruned a distilled LLM, deployed on a custom server with 5 RPS sustained throughput on a single A100 GPU, enabling low-latency streaming for B2B applications.
• Developed LLM-powered AI agents for automating e-commerce workflows, including SKU management and personalized recommendations based on client intent and prior chat history.
• Led fine-tuning of a vision-language OCR model for handwritten and printed invoices, enabling automatic SKU extraction and shopping basket population.
• Contributed to voice-to-text and text-to-voice model pipelines by designing data monitoring systems and fine-tuning processes to ensure dataset quality.

Scalable Data Collection & Processing Pipeline for LLM Training

Engineered a high-throughput data ingestion pipeline for LLM training that performs online quality control, tokenization, and Parquet sharding. It leverages Hugging Face Datasets (streaming), asyncio, and PyArrow for efficient ingestion with robust checkpointing.

AI Science & Deep Learning Implementations

• Implemented and benchmarked generative models from scratch (VAEs, GANs, DDPMs), quantifying the superior image generation of diffusion models.
• Patched the RoBERTa repository to add adapter support, achieving faster convergence and higher training accuracy.
• Engineered an image captioning system by integrating ViT and DistilGPT-2, and re-implemented underlying transformer modules from scratch.
• Implemented BYOL self-supervised learning, verifying improved classification accuracy on downstream tasks after pretraining.
• Conducted zero-shot object recognition experiments using OpenAI's CLIP model.

Ph.D. in Artificial Intelligence and Data Science

Thesis: Building and enhancing AI agents through Neuro-Symbolic and physics-informed machine learning.

M.Sc. in Scientific Machine Learning

Research-based Master's focused on the mutual application of computer science and engineering principles.

B.Sc. in Engineering

Specialization in Robotics.