Sharoz Haseeb
497 G1 JoharTown, Lahore, Pakistan
-
#-
ï LinkedIn
§ Github HuggingFace
Education
FAST National University of Computer and Emerging Sciences
Masters of Science in Computer Science
Forman Christian College University
Bachelor of Science in Computer Science
Jan. 2022 – Jan 2024
Lahore, Pakistan
May. 2017 – Oct 2021
Lahore, Pakistan
Experience
Kodezi
August 2021 – Ongoing
Founding Machine Learning Engineer [Remote]
Lahore, Pakistan
• Created POCs which helped raise $40,000 in Angel Investment and $800,000 in Pre-Seed Round
• Serverlessly deployed sophisticated pipelines with Finetuned OpenSource and Closed Source LLMs.
• Managed and worked with a distributed team over 4 countries.
Ekkel AI
January 2020 – January 2024
Artificial Intelligence Engineer [Onsite]
Lahore, Pakistan
• Build AI models from scratch and help the different components of the organization (such as product managers and
stakeholders) understand the results.
• Build data ingestion and data transformation infrastructure.
• Build APIs and handled deployment of AI infrastructure.
Projects
Kodezi Enterprise | Production Project | Deep Learning, AWS, SDLC
• Developed AI infrastructure for a platform which offers programmer productivity tools.
• Interacted with tech companies understanding their bottlenecks and offering solutions for them.
• Created POCs and MVPs for onboarding different tech enterprises, solving their bottlenecks.
MetaVox | Production Project | Audio Processing, Machine Learning, GCP
• Created sophisticated audio ingestion pipelines.
• Created an interactive Multi-Ai-Agent pipeline which performs edits in audio.
Sept 2021- Ongoing
Mar 2023 - Jan 2024
Lunchbreak AI | Production Project | Natural Language Processing, Language Modeling, AWS
Jun 2023 - Oct 2023
• Created a pipeline which used LLMs to generate long form content.
• Developed a data ingestion system which fed verified resources to the AI to reduce hallucinations.
• Built a system which utilizes the properties of perplexity and burstiness in text which incorporates humanizing elements
in text making it less prone to being flagged as AI generated.
Twinly AI | Production Project | NLP, In-Context Finetuning, Serverless, RAG
Oct 2023 - Ongoing
• Created a pipeline to develop prompts for chatbots according to user’s needs on the runtime increasing positive
engagement upto 60%.
• Developed a benchmarking system using Ai Agenst which adds new contextual examples into the Chatbots context to
tailor it rocketing conversion rates.
• Deployed the infrastructure to Aws Lambda for serverless serving and managed an EC2 as a middleware.
Profile Photos AI | Production Project | CV, Diffusion Models, Low-Rank, LCM, ETL, RunPod
• Created an ETL pipeline to Finetune Stable Diffusion v1.5
• Implemented Latent Consistency Modeling pipelines to reduce inference time by 90%.
• Curated a set of Low-Ranks weights which can be hot-swapped on inference time.
Sept 2023 - Dec 2023
AV Society | Production Project | CV, GANs, AWS
Mar 2022 - June 2022
• AV Society is a new an attractive way of generating NFTs. Proprietary AI models develop unique NFTs based on users
image. User can find and trade everything from rare digital art and branded crypto-collectibles, to digital fashion and
gaming items to use in multiple worlds.
• Researched and implemented pipelines of State of the Art GANs.
• Created serving APIs for multiple Pix2Pix image translation models.
Optimizing Code with LLMs: an Agent based Approach | Research Project | LLM
Feb 2023 - Dec 2023
• Researched the current gaps in Code Optimizations.
• Implemented multiple Ai Agents which worked in a simulated environment to optimize code bases.
• Achieved 50% jump in accuracy compared to vanilla systems.
AI-Driven Image Annotation for Object Detection | Research Project | ViT, Resnet
Sep 2023 - Nov 2023
• Created an approach which uses a Vision Transformer and a Zero Shot Object Detection model in coherence.
• Given a text prompt and a dataset the pipeline returns bounding boxes and segmented masks.
• Developed a pipeline which converts the coordinates to Pascal VOC XML so people can easily create HuggingFace and
Roboflow datasets.
GAN based Facial Editing of Real Videos | Freelance Project | GANs, GCP
Oct 2022 - Dec 2022
• Developed the data pipeline to feed the frame data of videos to deployed models.
• Fine tuned for custom data set, performed inversion, naive editing and stitch tuning of the edited frames.
• Features included gender swap, age manipulation, changing facial expressions and adding cosmetic layer to subjects face.
Urdu to Roman Urdu Translation | Freelance Project | GRU, LSTM, Transformer, Selenium
Oct 2022 - Jan 2023
• Fined-tuned GRUs, LSTMs, mT5, T5 base, GPT-J and other opensource LLMs for translation tasks.
• Sraped content from message boards, twitter etc to curate a dataset.
• Created a Gradio GUI for presentation purposes.
WriteNgo | Production Project | OpenAI, Lambda
Feb 2023 - May 2023
• Offers original quality content in seconds. Supervised a team to create multiple sophisticated Prompt Engineered
Pipelines which were deployed using AWS Lambda to be served as a server-less endpoint for to auto scale respective to
the number of users using the platform.
Sign Language Detection | Production Project | Real-Time, TensorFlow
April 2022 - May 2022
• Trained Single Shot Detection model on a custom data-set which was labeled using labelImg.
• Achieved 93% testing accuracy which helps detect signs in real-time, used tf-records to keep track of training.
Relevant Coursework
•
•
•
Deep Learning
Machine Learning
Natural Language Processing
•
•
•
Medical Image Processing
Algorithms Analysis
Operating Systems
•
•
•
Software Methodology
Theory of Computation
Research Methodology
Technical Skills
Languages: Python
Development Skills: Machine learning algorithms, Deep learning algorithms, Reinforcement learning, Natural language
processing (NLP), Computer vision, Data analysis and visualization, ML Optimization techniques, Model deployment and
maintenance, Agile project management, HuggingFace, Wandb, Google Cloud Platform, Amazon Web Services, RunPod
and Replicate
Technologies/Frameworks: Pytorch, TensorFlow, Linux, Jenkins, GitHub, ETL and RAG