Eve - Batch-Computing Runtime

Engineered a Python batch-computing runtime foundation across 10 core modules, integrating task scheduling, process execution, state tracking, and structured result handling. Designed deterministic FIFO scheduling, immutable execution models, and a 7-state lifecycle; validated behavior with 324 passing unit and integration tests. Implemented spawned-process execution for 2 allowlisted workloads, preserving identifiers across serialization and distinguishing structured workload errors from infrastructure exceptions.

GPT-Style Transformer Language Model

Implemented and trained a 14.3M-parameter decoder-only Transformer in PyTorch with 8 layers and 8 attention heads, building causal self-attention, positional embeddings, residual connections, and autoregressive text generation. Trained for 3,000 iterations on a character-level Wizard of Oz corpus, reducing validation cross-entropy from 4.46 to 1.63 using AdamW optimization and PyTorch CUDA acceleration.

Real-Time Pool Drowning Detection

Built a real-time pose-estimation pipeline that converts webcam streams into 0-100 drowning-risk scores using landmark geometry, limb visibility, verticality, and motion stasis. Added 3 false-alert safeguards - sustained-alert logic, pool-zone filtering, and multi-person tracking - and validated behavior against deck activity and out-of-zone movement.

✅ Verified by Extern · ⏱️ In progress

Pfizer Advanced: AI-Powered Document Insights & Data Extraction Externship

Prototype AI-powered document intelligence with Pfizer—using OCR, LLMs, and RAG to automate real enterprise PDF workflows and build a standout portfolio project.