Husnain Sattar

Academic

Verbatim text

Husnain Sattar
+92 320 7141398, sattar.husnain123@gmail.com
Gll, Islamabad
Linkedln: https://www.linkedin.com/in/husnainsr/
eep Learning, Data Analytics, Data Visualization, Datamining, Machine Learning, Advanced
tatistics and Reinforcement Learning
inal Pro·ect: Mitigating Financial Fraud via Deepfake Video Detection - Optimized vision pipelines (CNNs, GANs)
or real-time inference. [Deep learning, Computer Vision, Flutter]
eveloping advanced computer vision solutions for deepfake video detection, focusing on optimizing pipelines for
ear real-time erformance.
olBacha Chatbot: Designed and implemented an RAG (Retrieval-Augmented Generation) chatbot using a local LLM, a
ector database, and semantic caching for 2,000 daily users.
ools Used: LangGraph, Chroma, Streamlit
rames Generation using CV: Developed a video prediction framework using deep learning models (ConvLSTM,
redRNN, Transformer) to predict future video frames, capturing spatial and temporal dependencies effectively.
ools Used: PyTorch, HSPy, wandb, AWS Azure
roduct Recommendation for Amazon: Developed a ETL pipeline and application to efficiently recommend (128GB+)
roducts, using Spark and Hadoop for a seamless user experience.
ools Used: Spark, Kafka, Mongodb, streamlit
CV Research Assistant, Aim's Lab - Islamabad, PK
• Developed deepfake detection models using CNNs and GANs, achieving 10% higher
ccuracy than baseline on FaceForensics++ dataset.
Optimized face recognition pipelines (OpenCV, dlib) for real-time performance (~SOms
atenc ) in mobile bankin a
s (iOS/Android).
I Intern, AIO - Islamabad, PK
Engineered NLP-CV hybrid models for multimodal topic classification, improving accuracy
95%.
L Intern, Aim's Lab - Islamabad, PK
Built a CV-driven Saas platform for automated cricket highlights (YOLO, OpenPose),
chieving 97% accuracy in event detection.
kills & Tools
June 2024 - Present
Jan 2024 - March 2024
June 2023 -August 2023
ython, LangChain, OpenAI API, Chroma DB, Stream lit, PyTorch, HS Py, WandB, AWS, Azure, Flask,
Qlite, Chart.js, scikit-learn, bcrypt, smtplib, YOLO, OpenPose, TensorFlow, OpenCV, NumPy,
chievements
10 hackathon - 1st Place, AIO Hackathon 2024: Won out of 60 teams for a CV-NLP hybrid fraud detection solution.
cademic Performance - Published Research Paper, 2x Silver Medalist and Achieved continuous placement on the
ean's Honor List for 6 semesters in reco nition of outstandin academic erformance.
FAST NUCES ISLAMABAD CAMPUS
Ism a
+923065945261, ism a .h ussain 221@gma i I .com
Chakri Road, Rawalpindi
Linkedln: https://www.linkedin.com/in/isma-ahsan/
achelor of Science (Data Science)
eep Learning, Data Analytics, Data Visualization, Datamining, Machine Learning, Advanced
tatistics and Reinforcement Learning
inal Pro 'ect; Mitigating Financial Fraud in the Age of Deepfake Technology: Developing Robust Detection and
revention Methods [Deep learning, Computer Vision, Flutter]
eveloping advanced solutions for deepfake video detection, focusing on optimizing pipelines for near real-time
erforma nee.
hatify- Custom Chatbot using RAG: Developed a chatbot with OpenAI, LangChain, and ChromaDB, integrating
emantic search and vector databases for efficient document retrieval.
rames Generation using Computer Vision: Built a deep learning-based video prediction framework using ConvLSTM,
redRNN, and Transformers to capture spatial and temporal dependencies.
omprehensive Forecasting System: Developed an interactive forecasting system with Flask backend, Chart.js for
isualization, and SQLite for data management.
ecureAuth- Two-Factor Authentication Web App: Designed a secure authentication system using Flask, bcrypt for
assword hashing, and smtplib for email-based verification.
uman Exercise Recognition: Implemented real-time exercise recognition with YOLO for object detection, 3D pose
stimation, and CNN-based classifiers.
mage Compression and Reconstruction: Built a custom neural network for efficient image compression and
econstruction, optimized for low-latency edge devices.
light Delay Time Prediction: Created an ensemble learning model for predicting flight delays, leveraging feature
ngineering and hyperparameter tuning.
ill-in-the-Blank Prediction using LSTM: Developed a deep learning-based text prediction model using LSTMs, trained
n lar e text datasets for enhanced contextual accurac .
Undergraduate Research Assistant, AIM's Lab
• Developing advanced solutions for deepfake video detection, focusing on optimizing
ipelines for near real-time performance.
Designing lightweight machine learning models tailored for deployment on mobile
latforms, including iOS and Android .
kills & Tools
rofessional Skills
eadershi , Inter ersonal Skills, Communication.
June 2024 - Present
ython, LangChain, OpenAI API, Chroma DB, Streamlit, PyTorch, HSPy, WandB, AWS, Azure, Flask,
Qlite, Chart.js, sci kit-learn, bcrypt, smtplib, YOLO, Open Pose, TensorFlow, OpenCV, NumPy,
GBoost, pandas, Keras, NLP.
10 hackathon - Won where over 60 teams participated, and secured the pool prize of 2 lac.
ean's Honor List - Achieved placement on the Dean's Honor List in 7 semesters in recognition of outstanding
FAST NUCES ISLAMABAD CAMPUS

AI enrichment