Bhanu Harsha Y
Fueled by Espresso & Curiosity ☕
|
MS Computer Science @ Northeastern University • AI/ML & Full-Stack Developer
Projects
GPA
Publications
About
Graduate student at Northeastern University pursuing MS in Computer Science with a focus on Machine Learning, Generative AI, and MLOps. Published researcher with hands-on experience building production-ready AI systems and full-stack applications.
I specialize in developing intelligent systems that solve real-world problems, from NLP chatbots and time-series forecasting models to scalable web applications. AWS AI Practitioner certified with expertise in LLMs, RAG pipelines, and modern web technologies.
Education
Northeastern University
MS Computer Science • GPA: 3.8/4.0
Jan 2025 – May 2027
GITAM University
BTech Computer Science • 8.73/10 (Distinction)
Aug 2020 – May 2024
Certifications & Awards
- • AWS AI Practitioner Certified
- • Best Project - BACKUS100 ICCS 2024
- • Published at ICETES 2023 (AIP Proceedings)
- • ML with Python (IBM)
Experience
Graduate Research Assistant
InsightX Lab, Northeastern University
Jan 2026 – Present
Boston, MA
- → Developing a multi-objective prompt optimization system using DSPy to generate empathetic, factual, and non-toxic counter-narratives to online hate speech.
- → Designed a 16-metric reward framework across 6 categories (safety, tone, refutation, alignment, grounding, language quality) integrating Perspective API, MNLI, SBERT, BERTScore, and GPT-2 for automated evaluation.
- → Proposed a novel Socratic Engagement metric grounded in persuasion psychology research to encourage reflective questioning in counter-speech, contributing to the lab's submission pipeline.
- → Collaborating with faculty on publications targeting top-tier AI/ML conferences
Machine Learning & Data Science Intern
Fluentgrid Limited
May – Jul 2023
Visakhapatnam, India
- → Engineered production-ready NLP chatbot using Python, Keras, TensorFlow, and NLTK to automate client engagement with intent classification across 44 categories
- → Automated large-scale data ingestion and preprocessing pipelines using Python and tabula-py, reducing manual effort by 95%
- → Designed end-to-end data pipeline utilizing Selenium for web scraping and automated data acquisition from primary sources
Projects
AutoMend is an event-driven remediation platform that eliminates MLOps "Alert Fatigue" by transforming passive monitoring into active resolution. By bridging the gap between observability tools and infrastructure actions, it autonomously manages scaling, rollbacks, and retraining workflows through a generative, human-in-the-loop engine.
•Built distributed data pipeline across 6 datasets with Apache Airflow and Ray, producing 5,090+ ChatML training records and 3,033 anomaly sequences; trains LogBERT to classify 7 infrastructure anomaly types (GPU OOM, drift, pipeline stalls).
•Leverages a fine-tuned Llama-3 model to translate natural language prompts into executable, schema-validated JSON DAGs for rapid incident response.
•A no-code, drag-and-drop editor for engineers to define complex logic using pre-built "Tool Nodes" (e.g., Scale, Rollback, Switch Traffic).
•Integrated "Human-in-the-Loop" validation nodes that pause high-risk operations for manual approval via Slack
AI/ML-based system optimizing traffic flow and prioritizing emergency vehicles through real-time inference and dynamic signal adjustments.
•Developed predictive ML model using YOLO object detection for real-time traffic optimization
•Implemented emergency vehicle prioritization with computer vision classification
•Led team of 4; Best Project at BACKUS100 ICCS 2024
Full-stack Learning Management System with role-based access control, course management, and comprehensive quiz functionality.
•Built RBAC for 4 user roles with complete CRUD for courses, modules, assignments
•Developed quiz module with timed assessments, auto-submission, and automated scoring
•Deployed 43 RESTful APIs across 7 MongoDB collections on Vercel/Render
ML pipeline forecasting daily retail sales across 54 stores and 33 product families, processing 3M+ transaction records.
•Engineered temporal features with Fourier series and rolling window statistics
•Implemented stacked ensemble with Ridge Regression meta-learner
•Achieved Top 20% on Kaggle leaderboard
Full-featured virtual calendar with MVC architecture, design patterns, and multiple interfaces.
•Implemented Command, Strategy, Factory patterns with GUI/console/headless modes
•Multi-calendar support with timezone conversion and conflict detection
•60+ JUnit test cases covering edge cases and integration scenarios
End-to-end ETL pipeline with normalized schemas and star schema data warehouses for OLTP/OLAP systems.
•Designed 3NF relational schemas and star schema architectures
•Processed 140K+ records with automated reporting and stored procedures
•Integrated multiple data sources into cloud-hosted MySQL
Production-ready NLP chatbot for automated client engagement with intent classification across 44 categories.
•Designed sequential neural network for intent classification
•Integrated with JSON-based knowledge base for response handling
•Reduced manual processing time by 95% with automated pipelines
Skills
Languages
Web Technologies
ML/AI Frameworks
LLM & GenAI
MLOps & Data
Databases
DevOps & Cloud
Dev Tools
Let's Connect
Seeking Software Engineer & ML co-op opportunities for 2026. Let's build something impactful together.