Industry Experience

Developing a RAG pipeline leveraging BM25S retrieval and Qwen2.5-7B-Instruct over 2K+ pharmaceutical documents, integrating data from PostgreSQL and AWS S3

Deployed a Streamlit application for the QA Validation and Operations Team using TabPFN Regressor and Classifier models, engineering 40 trend-based features from 500K+ GMP manufacturing records to predict equipment breakdowns, achieving a 67.8% ROC-AUC

• Selected as 1 of 5 from 200+ applicants to extend DeepChem’s support for large-scale molecular language models.
• Extended the HuggingFaceModel class by implementing the generate() API, implementing batched autoregressive inference,tokenizer padding, and device-aware execution.
• Developed an OLMo wrapper class supporting regression, classification, multitask learning, and causal language modeling with custom prediction heads, pretrained checkpoint loading, mixed-precision inference, LoRA/QLoRA, and multi-GPU support.

• Selected for the competitive AI4ALL Ignite Summer 2026 Accelerator.
• Developed an end-to-end Braille-to-speech computer vision pipeline integrating YOLO11n, EfficientNet-B0, Liblouis, and Piper TTS, achieving 96.15% detection mAP50, 99.41% classification accuracy, and 97.32% character coverage.
• Trained and evaluated detection and classification models across 4 datasets and 5 model families, scaling to 10,117 augmented images and 193K+ character crops and performing end-to-end error analysis to identify false positive detection and reading-order failures.