ALGO LOG
MS Data Science · UMD · GPA 3.74

Ajaykumar
Balakannan

>

Data Scientist with production experience applying supervised ML, NLP, and anomaly detection on 15M+ records across clinical healthcare and workforce analytics. Drove 25–35% efficiency gains through predictive modeling and BI dashboards adopted by clinical directors and operational stakeholders.

Live K-Means · click canvas to resetinitialising…
Gradient descent · loss surfacewaiting…
Logistic classifier · live trainingwaiting…
3.74
GPA @ UMD
3+
Data Roles
15M+
Records Processed
99.5%
Pipeline Uptime
Experience

Work History



Sep 2024 – May 2026
Counseling Center, UMD
College Park, MD
Data Scientist
  • Engineered Python and SQL pipelines to ingest 10 years of clinical appointment records into analytics-ready formats, enabling downstream no-show prediction models and operational KPI reporting.
  • Developed supervised ML models (Logistic Regression, Random Forest) on EHR-style appointment datasets to predict no-shows, improving resource utilization by 25% and enabling data-driven operational scheduling decisions.
  • Applied NLP (TF-IDF, sentiment analysis) to classify 4,000+ student feedback responses, surfacing recurring service quality patterns that informed program redesigns adopted by UMD counseling staff.
  • Built Tableau and Power BI dashboards tracking utilization, cancellation, and engagement KPIs, adopted by UMD Counseling Center clinical directors for weekly operational reviews and resource planning decisions.
PythonSQLLogistic RegressionRandom ForestNLP · TF-IDFTableauPower BI
May – Aug 2025
Canaria Inc.
New York, NY
Data Science Intern
  • Built Python and SQL ingestion pipelines across 3 cloud systems (ClickHouse, PostgreSQL, AWS S3) on distributed job market datasets, ensuring data quality and integrity for downstream analytics workflows.
  • Built a Gradient Boosting salary prediction model on 40K+ US zip codes, engineering SOC code, seniority, and company-level features to enable real-time compensation forecasting via an auto-updating ClickHouse pipeline.
  • Implemented anomaly detection systems (Isolation Forest, HBOS, KNN, SVM) on 15M+ distributed records to validate data quality and integrity, reducing inconsistencies by 35% and improving pipeline reliability.
ClickHousePostgreSQLAWS S3Gradient BoostingIsolation ForestHBOS
Mar 2023 – Jul 2024
AastraZen Technologies
Chennai, India
Data & Analytics Engineer
  • Migrated a legacy client database to a hybrid MongoDB and PostgreSQL architecture with dbt-managed transformation layers, improving query throughput by 40% and enabling real-time reporting for a mid-market client.
  • Fine-tuned a Hugging Face DistilBERT model for automated support-ticket categorization and built a semantic search feature using a vector database, cutting manual ticket triage time by 45% across client support workflows.
  • Architected a real-time data streaming pipeline using Apache Kafka feeding a centralized AWS S3 data lake, supporting concurrent analytics workloads across 5+ client accounts with 99.5% uptime.
MongoDBPostgreSQLdbtDistilBERTVector SearchApache KafkaAWS S3
Skills

Technical Stack



Core Proficiency
Python · Pandas · NumPy · Sklearn92%
SQL Querying & Data Warehousing88%
ML / Anomaly Detection85%
Visualization · Tableau · Power BI82%
Cloud & Data Platforms · AWS76%
NLP · Deep Learning74%
Machine Learning
Logistic RegressionRandom ForestGradient BoostingXGBoostARIMAIsolation ForestSVM
Programming & Data
PythonSQLRPandasNumPyScikit-learn
Data Platforms & Warehousing
SnowflakeClickHousePostgreSQLRedshiftDatabricksAWS (S3, EC2)
Visualization & BI
TableauPower BIQlik SenseSigmaExcel (VLOOKUP/Pivot)
Analysis & Process
KPI TrackingRoot Cause AnalysisWorkflow AutomationAd Hoc Reporting
Projects

Research Projects



01 — COMPUTER VISION
Safeguarding Agricultural Lands from Animal Intrusion

YOLOv5 + CNN ensemble with automated data reconciliation system and hardware-integrated prototype for real-world deterrent responses in precision agriculture.

Accuracy: 90%
False Pos. ↓: 30%
Reliability ↑: 25%
YOLOv5CNNComputer VisionIoTHardware Integration
🏆 2nd Place · National Hackathon (500+ teams) · BIOGECKO Published
View on GitHub ↗
02 — TIME SERIES FORECASTING
Bitcoin Price Forecasting Pipeline

End-to-end forecasting pipeline with real-time Bitcoin data collection, ARIMA and XGBoost models, automated ETL workflows, and interactive Qlik Sense dashboards.

Accuracy: 85%
Manual Effort ↓: 40%
Real-time: Yes
ARIMAXGBoostTime SeriesQlik SensePythonSQL
View on GitHub ↗
03 — DATA ANALYTICS
Sage Ventures: Multifamily Analytics

Python and SQL analytics pipeline for a multifamily portfolio, quantifying loss-to-lease exposure and occupancy gaps across a relational schema of properties, units, and rent rolls. Automated Excel and Power BI reporting for portfolio stakeholders.

Loss-to-Lease: $2.1M+
Occupancy Gap: 26pt
SLA Tickets: 533
PythonSQLPower BIExcel AutomationReal Estate
View on GitHub ↗
Education

Academic Background



Master's in Data Science
University of Maryland
May 2026 · Maryland, MD
GPA: 3.74 / 4.0
Natural Language Processing · Machine Learning · Big Data Systems · Deep Learning
B.Tech — Electrical & Electronics Engineering
Sri Krishna College of Engineering & Technology
Aug 2020 – May 2024 · Coimbatore, India
CGPA: 8.4 / 10.0
Microcontrollers · Power System Analysis · Python · Machine Learning in Energy Systems
Let's Build
Together.

Open to full-time roles, internships, and research collaborations in data science, ML engineering, and analytics.