Hi, I'm kumkum!
AI Integrations Engineer | Cognitive Science Researcher
About me
I build practical AI products that move from prototype to production.
RoleAI Integrations Engineer
FocusLLMs + Automation
ResearchMemory + Emotion
LocationToronto
I am currently an AI Integrations Engineer at Northbridge, where I design and build AI-driven tools that automate and enhance business processes. My work ranges from rapid prototypes to production-ready systems integrating large language models (LLMs) and enterprise data.
Here are some technologies I have been working with recently:
PythonSQLJavaScriptCSS
Outside of work, I explore how emotions influence memory and frequency effects, combining cognitive science and data modeling. I am drawn to projects that connect technical precision with psychological insight.
Experience
AI Solutions Engineer
AUG 2025 – PRESENT
▷
Exploring AI solutions and integrating them into Northbridge systems to enhance automation and operational efficiency.
▷
Designing and deploying new tools and features to streamline data workflows and improve decision-making processes.
▷
Build and deploy LLM-integrated systems, from early-stage prototypes to production-ready products.
▷
Research and test new AI and ML models to evaluate performance, scalability, and business impact.
Data Engineer (Internship)
JUN 2024 – DEC 2024
▷
Developed an AI-powered chatbot using Retrieval-Augmented Generation (RAG) to translate natural language into SQL queries.
▷
Boosted performance by 40% via prompt engineering with open-source LLMs and Cortex API for contextual interactions.
▷
Automated data workflows using Snowflake and Power BI, delivering dashboards.
▷
Presented the chatbot's value at a company-wide town hall and collaborated with Snowflake's Cortex team to improve the API.
Education
Bachelor of Science in Computer Science, Cognitive Science, and Mathematics@ University of Toronto
Relevant Coursework
Data Structures & AlgorithmsSoftware EngineeringDatabase SystemsMachine LearningComputer NetworksOperating Systems
Projects
Concept Learning Model
Developed a cognitive model for concept learning using the Clarion architecture, explaining human concept processing and categorization by integrating sensory and linguistic representations through an associative merging mechanism.
Recipe Search Software
Developed a recipe search application in Java with a collaborative team, integrating APIs and user-driven contributions to boost daily active users by 40%.
Image Filter Server
Built a C++ server application enabling users to upload photos and apply filters via a user-friendly web interface, optimizing the image processing workflow for enhanced performance.
Item Response Theory (IRT) Model
Enhanced an IRT model for individualized learning assessments, achieving a 90% improvement in accuracy by optimizing model parameters and algorithms.
F1 Predictive Model
Developed a machine learning model using RNN and LSTM to predict F1 driver lap times and positions, achieving 80% accuracy and predictions within 3 seconds of target lap times.
DRM Sleep Simulation
Designed and implemented a Python-based simulation model utilizing LSTM neural networks to analyze the effects of sleep on DRM false memory formation, successfully replicating empirical study findings.
Contacts Tracking App
Designed and developed a contact lens tracking app using SwiftUI, optimizing UI rendering and overall app usability.
Research Code
A collection of research-focused code and experiments spanning cognitive science and AI-oriented modeling workflows.
Skills
Languages
Python
Java
SQL
JavaScript
Swift
C++
Frameworks
React
SwiftUI
GraphQL
REST APIs
Tools
Git/GitHub
Azure Data Factory (ADF)
Power BI
Snowflake
Technologies & Libraries
NumPy
pandas
dimod
Streamlit
Machine Learning
PyTorch
TensorFlow
Scikit-learn
LLMs
RAG
RNNs
IRT
NLP
Unsupervised Learning
Model Optimization