Aniket Sharma

Working as a Generative AI Developer at CGI in Edmonton, Alberta, Canada, building applied generative AI solutions.

• Engineered a LangGraph-powered AI agent hosted on Cloud Functions that orchestrates Gemini chat, vector similarity search, and rule-based reasoning to deliver cited building-code answers in real time
• Developed a custom document processor that chunks building code documents, extracts hierarchical sections, and writes embedding vectors to BigQuery, enabling low-latency Graph RAG for the agent
• Authored and maintained comprehensive Terraform IaC for the entire stack, achieving one-command, environment-agnostic deployment on GCP

• Developed a generalized algorithm for selective unlearning in neural networks, enabling efficient removal of any subset of training data
• Designed a utility function to quantify and compare the influence of data subsets on neural network parameters
• Formulated an orthogonal gradient descent approach to update models on specific data subsets while preserving the rest

Edmonton, Alberta, Canada.

• CMPUT 200 - Ethics of Data Science and Artificial Intelligence (Fall 2024 and Winter 2025)
• CMPUT 267 - Basics of Machine Learning (Winter 2024)
• CMPUT 201 - Practical Programming Methodology (Fall 2023)

Collaborated with five Canadian AI startups on real-world ML and LLM challenges through Amii's Work Integrated Learning program.

• Implemented a novel technique to generate unbiased synthetic data across multiple protected attributes (MultiFairGAN)
• Performed an extensive bibliographic search into bias mitigation techniques and prior works in AI in Education (AIEd) tools
• Evaluated and compared fairness of Knowledge Tracing algorithm

• Engineered incremental data integration pipelines from Salesforce to a weakly consistent global file system
• Programmed scripts to facilitate data curation and analysis across different Alphabet organizations
• Constructed dashboards to provide key security and privacy insights about Salesforce footprints across Alphabet to the leadership and stakeholders

• Designed and implemented a simulation of human episodic memory (accepted in ICAAI 2022)
• Devised a data-centric approach in order to improve the performance of machine learning models predicting HCV severity by approx. 16% (accepted in ISDA 2021)
• Developed an open-source web crawler with quality features extractor for wikiHow (accepted in IATMSI 2022)

• Created a production-ready tool for solving Capacitated Vehicle Routing Problem with Time Window and Multiple Depots (CVRPTWMD) with additional constraints which reduced logistics costs by 20%
• Employed modified K-means clustering algorithm with automated elbow method for solving Dynamic Facility Location and Allocation Problem
• Hosted the developed tool using serverless microservices on AWS

• Wrote 50+ blogs with cumulative 60,000+ views on the topic of AI/ML and Python
• Created an Event Prediction Model using sentiment analysis of real-time tweets fetched using Twitter API
• Contributed to Fingerprint Recognition System and Motion Detection System