Collection of Independent Projects

Enhancing Chest X-Ray Image Interpretation Using a Vision Language Model (VLM)

August 2024 - December 2024
  • Created a unified framework with a new pipeline between existing AI medical report generation and a fine-tuned VLM, providing additional context for medical Q&A, resulting in higher-quality image interpretation.

  • Outperformed the vanilla open-source 1.8B Moondream 2 VLM by 23% in medical answer accuracy after finetuning it with the Visual Question Answering Radiology dataset and using the proposed framework.

  • Developed guardrails to decline unsafe medical questions based on a combined score generated by fusing individual safety evaluations from a group of three Microsoft Phi-3 Mini LLMs using Reciprocal Rank Fusion.


    [Paper] [Code]

Enhancing Chest X-Ray Image Interpretation Using a Vision Language Model (VLM)

August 2024 - December 2024
  • Created a unified framework with a new pipeline between existing AI medical report generation and a fine-tuned VLM, providing additional context for medical Q&A, resulting in higher-quality image interpretation.

  • Outperformed the vanilla open-source 1.8B Moondream 2 VLM by 23% in medical answer accuracy after finetuning it with the Visual Question Answering Radiology dataset and using the proposed framework.

  • Developed guardrails to decline unsafe medical questions based on a combined score generated by fusing individual safety evaluations from a group of three Microsoft Phi-3 Mini LLMs using Reciprocal Rank Fusion.


    [Paper] [Code]

Bag Theft Detection System with Real-Time Alert Notifications

November 2024 - December 2024
  • Prototyped a portable theft detection device designed to be placed in a user's bag using the ESP32 Feather microcontroller and Adafruit MPU-6050 6-DoF Accelerometer/Gyro sensor.

  • Utilized the If-This-Then-That (IFTTT) IoT Platform to interface with the Google Assistant mobile app, allowing the user to arm/disarm the system using voice commands.

  • Logged voice commands in the ThingSpeak IoT Platform using Webhooks, periodically polled for system state changes, and sent updates to ThingSpeak over Wi-Fi.

  • Illuminated various LEDs on the ESP32 and sent alert notifications to the user's phone if unusual motion was detected by the MPU-6050 sensor.


    [Demo Video]

Bag Theft Detection System with Real-Time Alert Notifications

November 2024 - December 2024
  • Prototyped a portable theft detection device designed to be placed in a user's bag using the ESP32 Feather microcontroller and Adafruit MPU-6050 6-DoF Accelerometer/Gyro sensor.

  • Utilized the If-This-Then-That (IFTTT) IoT Platform to interface with the Google Assistant mobile app, allowing the user to arm/disarm the system using voice commands.

  • Logged voice commands in the ThingSpeak IoT Platform using Webhooks, periodically polled for system state changes, and sent updates to ThingSpeak over Wi-Fi.

  • Illuminated various LEDs on the ESP32 and sent alert notifications to the user's phone if unusual motion was detected by the MPU-6050 sensor.


    [Demo Video]

Custom Networking Protocol for Reliable ASCII File Transfer

November 2024 - December 2024
  • Developed a custom OSI Transport Layer TCP-like protocol using Python to reliably transfer an ASCII file between a simulated client and server with a router in between.

  • Implemented sliding window and buffering algorithms to manage client/server receive/send buffers.

  • Guaranteed reliable file transfer using Cumulative and Selective Acknowledgements with Fast Retransmit and timeouts despite random packet drops.

  • Reinforced my understanding of TCP Congestion/Flow Control and the importance of designing algorithms that can efficiently and reliably transfer data with minimal latency and memory overhead.

Custom Networking Protocol for Reliable ASCII File Transfer

November 2024 - December 2024
  • Developed a custom OSI Transport Layer TCP-like protocol using Python to reliably transfer an ASCII file between a simulated client and server with a router in between.

  • Implemented sliding window and buffering algorithms to manage client/server receive/send buffers.

  • Guaranteed reliable file transfer using Cumulative and Selective Acknowledgements with Fast Retransmit and timeouts despite random packet drops.

  • Reinforced my understanding of TCP Congestion/Flow Control and the importance of designing algorithms that can efficiently and reliably transfer data with minimal latency and memory overhead.

Distributed Routing and Forwarding Algorithm for Packet-switched Networked Systems

October 2024 - November 2024
  • Implemented distributed Distance Vector routing using Python to facilitate optimal data packet transfer between simulated clients via routers configured in various network topologies.

  • Developed software to manage routing tables and created forwarding logic running in parallel at each router to parse incoming control packets.

  • Demonstrated resiliency to new link creations, sudden link failures, and other common networking issues such as loops and Count to Infinity by ensuring receipt of all transmitted data packets.

  • Reinforced my understanding of network routing concepts and gained practical experience implementing a fault-tolerant distributed algorithm, an essential part of real world networked systems.

Distributed Routing and Forwarding Algorithm for Packet-switched Networked Systems

October 2024 - November 2024
  • Implemented distributed Distance Vector routing using Python to facilitate optimal data packet transfer between simulated clients via routers configured in various network topologies.

  • Developed software to manage routing tables and created forwarding logic running in parallel at each router to parse incoming control packets.

  • Demonstrated resiliency to new link creations, sudden link failures, and other common networking issues such as loops and Count to Infinity by ensuring receipt of all transmitted data packets.

  • Reinforced my understanding of network routing concepts and gained practical experience implementing a fault-tolerant distributed algorithm, an essential part of real world networked systems.

Developing an HTTP Client-Server System in C for Various Content Retrieval

September 2024 - October 2024
  • Implemented an HTTP client instance that can reliably retrieve various content (web pages, zip files, images, PDFs, etc.) from an HTTP server.

  • Developed a simulated HTTP Content Delivery Network (CDN) server to deliver cached content to clients over TCP and retrieve additional items from an external database over UDP.

  • Implemented the client and server instances fully from scratch using socket programming in C and working at the bit-level to ensure corruption-free content retrieval.

  • Reinforced my understanding of fundamental networking protocols, efficient memory management, and data integrity verification techniques.

Developing an HTTP Client-Server System in C for Various Content Retrieval

September 2024 - October 2024
  • Implemented an HTTP client instance that can reliably retrieve various content (web pages, zip files, images, PDFs, etc.) from an HTTP server.

  • Developed a simulated HTTP Content Delivery Network (CDN) server to deliver cached content to clients over TCP and retrieve additional items from an external database over UDP.

  • Implemented the client and server instances fully from scratch using socket programming in C and working at the bit-level to ensure corruption-free content retrieval.

  • Reinforced my understanding of fundamental networking protocols, efficient memory management, and data integrity verification techniques.

Building an Audio Equalizer from Scratch

March 2023 - April 2023
  • Built an audio equalizer that can adjust the frequency levels of audio from a source connected via a 3.5mm headphone jack and play it through a 2-way speaker.

  • Created a system schematic in LT Spice and performed calculations and Frequency Response Analyses to select optimal components including resistors, capacitors, various amplifiers, other ICs, etc.

  • Constructed Bass, Mid, and Treble filters to adjust the amplitude in three frequency bands using three potentiometers.

  • Conducted an in-depth analysis of all system modules using signal generators, oscilloscopes, and other benchtop equipment to minimize voltage leakage and maximize the signal driven to the speaker output.


    [Demo Video]

Building an Audio Equalizer from Scratch

March 2023 - April 2023
  • Built an audio equalizer that can adjust the frequency levels of audio from a source connected via a 3.5mm headphone jack and play it through a 2-way speaker.

  • Created a system schematic in LT Spice and performed calculations and Frequency Response Analyses to select optimal components including resistors, capacitors, various amplifiers, other ICs, etc.

  • Constructed Bass, Mid, and Treble filters to adjust the amplitude in three frequency bands using three potentiometers.

  • Conducted an in-depth analysis of all system modules using signal generators, oscilloscopes, and other benchtop equipment to minimize voltage leakage and maximize the signal driven to the speaker output.


    [Demo Video]

Optimizing Search and Rescue Missions for Autonomous Vehicles with Drone Collaboration

January 2022 - May 2022
  • Simulated an off-road Search and Rescue mission in the Unity Game Engine with an autonomous ground vehicle (AGV) and a UAV.

  • Applied Deep Learning, Computer Vision, and Sensor Fusion (Camera + LiDAR + Depth Sensor) techniques to assist the AGV in performing terrain classification and obstacle detection/avoidance.

  • Utilized a UAV to perform aerial reconnaissance and intelligently re-route the AGV using visual information captured from its camera.


    [Project Presentation]

Optimizing Search and Rescue Missions for Autonomous Vehicles with Drone Collaboration

January 2022 - May 2022
  • Simulated an off-road Search and Rescue mission in the Unity Game Engine with an autonomous ground vehicle (AGV) and a UAV.

  • Applied Deep Learning, Computer Vision, and Sensor Fusion (Camera + LiDAR + Depth Sensor) techniques to assist the AGV in performing terrain classification and obstacle detection/avoidance.

  • Utilized a UAV to perform aerial reconnaissance and intelligently re-route the AGV using visual information captured from its camera.


    [Project Presentation]