Experience
I build reliable AI systems: LLM pipelines, robust ETL/crawling pipelines, and backend/data infrastructure that turn messy text into decisions.
💼 Professional Experience
University of Waterloo
Graduate Researcher (ML/NLP)
2022–2025, Waterloo, ON, Canada
- Reframed design cognition methods as NLP tasks and ran systematic, reproducible LLM experiments.
- Curated in-context learning examples, instructions, and preference datasets for conversational meeting analytics.
- Applied parameter-efficient fine-tuning (LoRA/QLoRA) via Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO) on open-weight LLMs, utilizing tools such as Transformers, TRL, and Unsloth.
- Built semantic-search RAG pipelines and performed context engineering with closed-source and open-weight LLMs to improve retrieval and grounding.
- Utilized vector databases (e.g., Qdrant, Pinecone, FAISS) and prompt monitoring tools (e.g., Opik); implemented advanced features such as cross-encoder reranking and self-query.
- Performed offline model comparisons via custom evaluation pipelines using LLM-as-a-judge and RAGAS.
- Presented an earlier component of my research at the 11th International Conference on Design Computing and Cognition (DCC) 2024, in Montreal, Canada.
Earlier Experience
Socio Cognitive Lab
Machine Learning Researcher (NLP)
2020–2022, Dhaka, Bangladesh
Conducted research in NLP, social media mining, and social computing, focusing on data collection, analysis, and ML model development.
- Developed machine learning models to predict insomnia from 4.2M+ tweets across 1.5K+ users, collected using an ETL pipeline (Tweepy, SQL), leveraging Sentence-BERT (SBERT), IBM Watson, TensorFlow, and scikit-learn. Published in JMIR [IF≈6], DOI: 10.2196/27613.
- Estimated changes in Big Five personality traits by building a custom ML model using the PSO algorithm on 75K+ comments from 275 Facebook users’ ego networks. Published in SBP-BRiMS (2021).
Data & ML Engineer (Consultant)
2019–2023
Worked with clients globally to deliver data-driven solutions across finance, e-commerce, and automation domains.
- Built and operated large-scale web crawling/ETL (Scrapy, Selenium) and automation pipelines (Realtor, Indeed, LinkedIn, FB Marketplace, Instagram, X (Twitter), PulsePoint) with monitoring, retries, and CI/CD for reliable multi-source ingestion.
- Integrated NLP/ML into ETL (classification, extraction, Retrieval-Augmented Generation (RAG)) and built dashboards for financial events (e.g., M&A) and property analytics.
- Orchestrated cloud jobs, scheduling, and infrastructure (Airflow, AWS Batch, Vultr APIs).
Ergo Ventures Limited
Software Engineer (Python, Backend)
2020, Dhaka, Bangladesh
- Built scalable backend systems for e-learning and e-commerce platforms.
- Performed automation, data migration, and modeling using the Python tech stack.
DrutoTech
Software Developer (Python, Backend)
2019, Dhaka, Bangladesh
- Designed, developed, and maintained ETL and web crawling pipelines and automation scripts.
- Developed an admin panel and REST APIs for an e-commerce platform.
🎓 Education & Honors
University of Waterloo
MASc in Management Science and Engineering, Sep 2022 – Dec 2024
Research Focus: Machine Learning, Natural Language Processing
Academic Scholarships: Fully funded; Graduate Research Studentship (GRS); International Master’s Award of Excellence (IMAE) 2022–2024.
Courses: Big Data Analytics (MSCI 623), Text Analytics (MSCI 641), Foundations of ML (MSCI 700), Statistical Methods for Data Analytics (MSCI 718)
American International University-Bangladesh
BSc in Software Engineering, 2019
Capstone: Seradam – an e-commerce metasearch platform (Django, React.js)
Scholarship: Awarded for academic performance
Selected courses: Data Warehousing & Data Mining, Statistics & Probability, AI & Expert Systems, Computer Vision & Pattern Recognition