Goal: develop pose-matching image generation models to generate photorealistic football players.
AI Sweden
AI for Impact
2024 Summer
Developing a RAG-based chatbot enabling efficient retrieval of trustworthy, socioeconomic using natural language. Built for the non-profit Reach For Change, funded by Google.org - Google's philanthropic arm.
Built/integrated backend infrastructure to React frontend, leveraging tools such as Docker, FastAPI, LangChain, AWS Lambda.
Versed myself in SOTA RAG designs, such as agentic RAG (LangGraph), GraphRAG, self-reflection, logical routing.
Axis Communications
R&D intern
2023 Summer
Researched/developed machine learning models for snow-detection to prevent view obstruction in surveillance cameras.
Building and evaluating various Generative Adverserial Network (GAN) models for synthetic data generation.
Extending binary snow classifier to multi-class classifier whilst reducing quantized model size by ~60% and increasing accuracy to >90% with hyperparameter tuning.
Lunicore
IT Consultant
2022 Fall
Supported clients with projects related to database migration, cost-analysis, database & bill management with Amazon Web Services.
Ericsson
R&D intern
2022 Summer
Developing UI features for analytical toolboxes and automating data flow of KPI analysis between new commits, mainly in Python.
Optimizing memory allocation of physical layer signals, parsed from log files, in virtual cellular hardware for simulations with C.
Societies
KTH AI Society
AI Developer
2024 October - present
Twiga - WhatsApp bot for Tanzanian teachers. Leveraging resources from Tanzanian Instutite of Education (TIE) to provide pedaogical guidance to teachers.
Exploring and implementing vision-language models, graph neural networks, RAG designs to improve document understanding for queries.
Lund Formula Student
Driverless Developer
2021-2022
Developing software pipeline and simulations for autonomous go-kart like car for university-based Formula Student team.
Building multi-modal sensor system with INS, LiDAR and two-cameras to create real-time perception and path-planning pipeline.
LINC
Quantitative Research Analyst
2020-2021
Back-testing harmonic trading strategies in currency markets in collaboration with AI-driven hedgefund Valid Alpha.
Massive (>1 million data-points) data processing with Numpy/Pandas, data visualization with matplotlib/seaborn libraries, automating significance testing, technical analysis. Resulted in published paper.
Scraped LinkedIn job postings for ML, extracted tokens corresponding to skills with JobBERT model. Continual fine-tuning with groundtruths generated with multi-shot learning & prompt engineering
Visualized embedded and clustered skills with Plotly, deployed on HuggingFace Spaces.