M.Asad Asrar Minhas

Academic

Verbatim text

ole
st
M.Asad Asrar Minhas
+92 332 5680738, i191731@nu.edu.pk, SCHS Shifa Housing Society Rawalpindi
LinkedIn: https://www.linkedin.com/in/asad-minhas-5a5a05250/
Github: https://github.com/AsadMinhas782
Medium: https://medium.com/@i191731
\Education
Bachelor of Science (Artificial Intelligence) 2023
Major: rtificial Intelligence, Artificial Neural Networks, Machine Learning, Deep Learning, Natural
Language Processing, Computer Vision, DevOps, Computer Networks, Knowledge Representation &
Reasoning and Software Project Management & Engineering.
Alevels (Physics, Chemistry, Mathematics)
Olevels (Physics, Chemistry, Maths)
Projects
Final Project: MockVue — A mock job interview system for students to practice interviews in the field of Al
‘A website through which students can practice giving Al job interviews, which will analyze both their answers and their
facial expressions. It will generate a score in the end showing how well the user performed during the interview.
‘Semester Projects:
- Bitcoin Price analysis using NLP
+ DB SCAN, Hill Climbing Search, K-Nearest Neighbors (KNN) algorithm and K-Mean Clustering in Python.
+ Customized encoder and decoder model for image segmentation in Python Jupiter.
- Customized architechture of VGG-16
- Football Ontology Design using GraphDB, SPQRQL and Protege
+ Spell Correction and Poetry Generation for Roman Urdu language in Python Jupiter.
- TIC-TAC-TOE agent with dynamic board size using minimax algorithm in Python.
- Created a database managemnet system for a fitness application using SQL, P-SQL, HTML, PHP and CSS.
- MLOps Model Pipeline
- Inter-process communication in Linux using Fork and Named Pipes.
Work Experience
Fiverr and Upwork: Multiple gigs set up including focused on python Jun 2022 - current
Data Engineer Intern - IKNEX lab: Data curation & feature engineering using python and sql Jun 2022 — Aug 2022
\Skills & Tools
Professional Skills Leadership, Communication, Decision Making, Teamwork, Analytical Skills
Technical Skills Python, C++, OOP, TensorFlow, Github, Jenkins, Docker, PL/SQL, MongoDB, Oracle, Protege,
HTML, Pytorch, OpenCV, NLP, CV, MlOps
Achievements
- Won 1° place in Social Media Smackdown Nascon 23
Trainings / Certification
-Machine Learning Specialization - Coursera
-Neural Networks and Deep Learning - Coursera
Activities
President at FULL SEND ATHLETICS (22-23)
Vice President at FULL SEND ATHLETICS 21-22
Secretary Marketing NaSCon ‘22
\Interests
Fitness, Football, Cricket, Reading, Video Games
FINAL YEAR PROJECTS
Oe x
3C — 3D Characterizer
3C is a desktop-based application that allows the user to create a 3D character using a single 360
video of person. The application can be easily installed from 3C website which gives and easy
installer using which person can easily install and user application. Application is designed to be
user friendly as much as possible and to give information related to each step inside. The
application is highly GPU sensitive as it required GPU for rendering of point cloud.
The character that is generated by 3C is easily importable inside gaming engines such as unreal
engine. The character can easily be rigged in blender and users can make different animation on
the character as well using blender. User can perform various steps such as:
User video and take frames from the video then perform foreground extraction.
User can also generate a cartonized character.
- User can then user NVidia’s instant-ngp to make point cloud then mesh.
- Then Texture is extracted of the model.
Thanks to Nvidia’s Instant-ngp using which this became possible. With instant-ngp, it became
possible to create a mesh from point could. Using different models, it was possible to generate a
character that can easily be importable inside unreal engine.
A desktop based application for generating 3D model of an object.
which is importable inside blender and unreal engine by using
360 video of object.
Background Removal.
Technology Used:
ee) Python, Blender, Nvidia, PyTorch, PyQT, HTML,
The model is loaded into the = UV Mapping 3D Model Generation.
qetcetlon iy ones CSS, JS, React, Unreal Engine
Supervisor Name:
Ms. Humera Sabir
Creating a 3D model using neural Importing character into unreal engine
radiance field and segmentation. and displaying its functionality Group Members:
Foreground extraction Desktop application & testing. Noman Asif (i119 os 1880)
for crisp 3D model In-Game demo,
Afaq Qureshi (i119 - 1775)
Faizan Zubair (i19 - 1863)
Ms. Humera Sabir
Noman Asif - (191-1880)
Afaq Qureshi - (191-1775)
Faizan Bin Zubair - (19i-1863)
ayn
he
Al based conversation assistant designed for farmers
KisaanDost is a Mobile-first Web Application/Android Application where farmers can have a phone
call-style conversation in which they query our chatbot to gain solutions to their problems. The
chatbot is implemented in Python language. First, the user message is used to perform both intent
classification and entity extraction. Intent Classification helps us judge which problem the user is
trying to solve. Entity Extraction helps us gather the information we need from the user to solve the
user’s problem. Then Response Generation takes in user messages, intents, and entities as input to
respond to the user. As speech is a natural medium for communication, the idea of a voice-powered
chat will allow easy querying of useful information. It would help enable many farmers to ask their
questions and get answers in our local language.
Features include:
e Bloomz LLM finetuned with local Agri-data sources
e Anaesthetically pleasing and visibly understandable UI/UX
e Factually backed dataset fed by industry professionals.
e Urdu voice to voice integration
e Location API with weather update integrated.
e Personal farmer user profile
2 A(R
ag AN AI-BASED
‘ag! CONVERSATION
ASSISTANT DESIGNED
KisaanDost FOR FARMERS.
Supervisor: Dr. Mirza Omer Beg
S.M Asad (191-1778) Danial Nasir (191-1861) Asfand Ali (191-1656)
mmc4om4-z0a>
Technology Used:
Python, Html, CSS, Javascript, Flask, Figma,
Pytorch, Visual Studio
Supervisor Name:
Dr. Mirza Omer Beg
Iteration 1
sept-oct
mz-rm3-4
Group Members:
Syed Muhammad Asad (i19 - 1778)
Danial Nasir (i119 - 1861)
Asfand Ali Irfan (119 - 1656)
EeBaeBb 2 @W
ple Se
DEEPFARM - Agricultural Information & Insights
Deepfarm is a web application that utilizes modern Al techniques, specifically deep neural
networks and machine learning models, to provide insights and statistics for the agricultural
sector. The application uses open-source satellite imagery (Sentinel 2) to analyze crop fields and
provide useful information for decision makers and other stakeholders.
One of the key features of Deepfarm is its ability to Identify Different Crops that are being grown
in a particular field. By analyzing satellite imagery, the application can distinguish between various
crops such as wheat, corn and potatoes, among others. This information can be extremely helpful
for stakeholders who need to keep track of their crops and make informed decisions about their
farming practices.
Another feature of Deepfarm is Crop Segmentation. This refers to the process of dividing a crop
field into smaller sections, which can be helpful for analyzing and managing different areas of the
field. The application can segment a field based on crop type.
Finally, Deepfarm offers Crop Yield Estimation, which is the process of predicting the amount of
crop that will be harvested from a particular field. By analyzing satellite imagery and other data,
the application can provide estimates of crop yields, which can be helpful for decision makers and
other stakeholders in planning import/export decisions and managing their harvests.
Overall, Deepfarm is an attempt for digitalization of the agricultural data which is a powerful tool
that combines modern Al techniques with satellite imagery to provide valuable insights and
statistics for the agricultural sector. By helping stakeholders to identify crops, segment fields, and
esti

AI enrichment