BETSALEEL HENRY

Proactive and goal-oriented professional with a strong emphasis on time management and problem-solving. Recognized for reliability and adaptability, demonstrating a quick ability to learn and apply new skills. Committed to utilizing these qualities to drive team success and make valuable contributions to organizational growth.

Custom Assistant: Llama model on Raspberry Pi

Embedded systems Project    09/2023– 01/2024

· Compressed the Llama2 7B model to 40% of its original size using activation-aware weight quantization (AWQ).

· Improved 70% accuracy on Winogrande downstream tasks with only 5% corpus via lightweight fine-tuning on an NVIDIA GeForce RTX 3090 GPU. 

· Deployed the compressed model on a Raspberry Pi  (2.4 GHz processor) for inference with the utilizing llama.cpp library; 

·  Wrote shell script and ssh migrations to handle prompt passed to the model, used Django backend for user interaction.

Electronic Circuit Node Analysis

Computer Aided Circuit Design and Automation Project  09/2022– 01/2023

· Developed an Electronic Design Automation (EDA) tool in  C++ that parsed circuit connections in a netlist.

· Achieved less than 10 Second latency, and support for components such as BJTs and MOSFETs using their small-signal models.

· Implemented Modified Nodal Analysis (MNA) to compute and generate comprehensive reports of voltages and currents.

Reinforcement Learning Virtual Kart

Embedded systems Project   02/2022– 09/2022

· Collaborated with two peers to develop a Deep Reinforcement Learning Agent on Jetson Nano, for autonomous navigation of a Kart in Python, based on continuous control mechanism, Proportional, Integral, Derivative (PID) tuning.

· I Dealt simultaneously with complex environmental dynamics and our complex policy, derived from streamed frame pixels.