Nishad Mansoor
data science @ northeastern open to work
data science @ northeastern open to work
Built a pipeline tracking AI usage across the firm using the Anthropic and OpenAI APIs, then a time-series model projecting $1.7M in year-end spend. Additionally, I built a conversational agent that sits on top, answering cost questions in plain language against a Power BI model on Azure.
Benchmarked four vision-language models across 300+ origami models using IoU and CLIP similarity to see how they understand spatial reasoning, tuning constraint thresholds to query efficiency scores of 0.85–0.87. Published the paper to workshops at ICML and CVPR.
Built an ingestion pipeline pulling DoD contract data from USASpending.gov and geocoding vendor locations, feeding a Neo4j knowledge graph that links vendors, contracts and geographies through UEI/CAGE identifiers.
Generated CSVs from parsing HTML docs and built an AI research dashboard that visualized trends from student reflections for use in future semesters.
Trained a random forest on marketing spend and 1,000 survey responses, reaching an F1 of 0.84 for audience targeting, and built a TF-IDF sentiment pipeline to surface what drove customer satisfaction.
