Taljinder Singh
ASU MS-DS '26 | 2+ Years Experience
Transforming raw data into actionable insights through machine learning, scalable cloud pipelines, and AI-driven solutions.
Transforming Data into Business Impact
Data scientist and ML engineer with a passion for building intelligent systems
My Journey in Data Science
I am a graduate student in Data Science at Arizona State University with 2+ years of professional experience in analytics and pricing strategy at OYO and Indus Insights.
My expertise spans machine learning, forecasting, optimization, and cloud-based data pipelines. I have a proven track record of designing scalable solutions that improved forecasting accuracy, optimized pricing models, and drove measurable revenue growth.
Core Competencies
Project Notebooks
Interactive data science workflows with code, visualizations, and insights
AI Indian Recipe Generator
Explored the use of Large Language Models (LLMs) for AI-driven recipe generation based on input ingredients.
# Fine-tuning LLMs for Recipe Generation from transformers import AutoModelForCausalLM, AutoTokenizer from peft import LoraConfig, get_peft_model import torch # Load base model (LLAMA3.2) model_name = "meta-llama/Llama-3.2-7B" tokenizer = AutoTokenizer.from_pretrained(model_name) # Configure QLoRA for efficient fine-tuning # Click "Run All" to see full project details
Book Recommender System
Developed a personalized book recommendation system using collaborative filtering techniques on the Book-Crossing dataset.
# Collaborative Filtering for Book Recommendations
import pandas as pd
import numpy as np
from sklearn.metrics.pairwise import cosine_similarity
# Load Book-Crossing dataset
ratings_df = pd.read_csv('BX-Book-Ratings.csv')
books_df = pd.read_csv('BX-Books.csv')
# Build user-item matrix for collaborative filtering
# Click "Run All" to see full project detailsHR Analytics Dashboard
Created a visual HR analytics dashboard using Tableau as part of coursework at Arizona State University.
# HR Analytics Dashboard - Data Preparation
import pandas as pd
import numpy as np
# Load HR dataset
hr_data = pd.read_csv('hr_dataset.csv')
# Analyze employee attrition patterns
attrition_rate = hr_data['Attrition'].value_counts(normalize=True)
# Prepare data for Tableau visualization
# Click "Run All" to see full project detailsYelp Arizona Analysis
Performed comprehensive analysis of the Yelp dataset for Arizona, focusing on restaurant businesses and user behavior.
# Yelp Arizona Analysis - PySpark Processing
from pyspark.sql import SparkSession
from pyspark.sql.functions import col, count, avg
# Initialize Spark session
spark = SparkSession.builder.appName("YelpAnalysis").getOrCreate()
# Load Yelp JSON files (business, review, user, tip, checkin)
business_df = spark.read.json("yelp_academic_dataset_business.json")
# Filter for Arizona restaurants
az_restaurants = business_df.filter(col("state") == "AZ")
# Click "Run All" to see full project detailsProfessional Projects
Real-world solutions delivering measurable business results
Brent & Gas-Oil Spread Forecasting
Challenge
Energy trading desks rely on accurate short- to medium-term forecasts of commodity spreads for daily trading decisions, hedging strategies, and portfolio rebalancing. Existing Excel-based statistical models lacked the ability to capture non-linear dependencies between macro variables and price spreads, adaptability to shifting volatility regimes during global market fluctuations (e.g., OPEC decisions, inflation shocks), and automated performance monitoring.
Solution
Developed advanced LSTM-based forecasting framework to predict the spread between Brent crude oil and Gas-Oil prices. Aggregated 10+ years of daily data from multiple sources (EIA, Bloomberg, proprietary feeds). Engineered 30+ lag and trend features using temporal shifting, moving averages, and differencing. Implemented LSTM architecture in TensorFlow/Keras with grid-search tuning for sequence length, hidden units, dropout, and learning rate. Used walk-forward cross-validation to simulate production deployment and avoid look-ahead bias. Added model-drift monitoring pipeline for automatic retraining triggers.
Price & NPV Prediction Models
Challenge
Financial analysts needed standardized NPV estimation across verticals, accounting for real-time macro changes (inflation, interest rates, FX rates) while reducing human bias in manual spreadsheet forecasts. The client's decision pipeline lacked automation and cross-portfolio comparability, leading to inconsistent pricing strategies and delayed approvals.
Solution
Developed a comprehensive pricing and NPV prediction framework integrating macroeconomic indicators and deal-specific features. Engineered 20+ predictive features capturing macroeconomic trends (interest rate spread, CPI, PMI), industry-specific multipliers, and deal-level variables. Built feature store for reusable variables across business units. Implemented regularized regression (Ridge, Lasso) and tree-based ensembles (XGBoost, RandomForest) with cross-validated grid search. Automated pipeline for quarterly model retraining and deployed as Python microservice with REST API endpoints for real-time price recommendations.
Revenue Optimization Framework
Challenge
The existing pricing system at OYO relied heavily on manual analyst decisions, leading to inconsistent pricing across markets and property types, inability to adapt quickly to demand shocks (festivals, events, weather), and limited visibility into pricing rationale for on-ground teams.
Solution
Designed comprehensive pricing and revenue optimization framework for 100+ hotel properties. Integrated data from multiple sources: booking pace, competitor prices, search trends, and event calendars. Implemented ensemble forecasting pipeline combining ARIMA, Prophet, and Gradient Boosting Regressor to forecast occupancy and ADR at property and cluster level. Designed price optimization module using mathematical programming (SciPy's optimize.minimize) to maximize revenue subject to occupancy, min/max price bounds, and competitor constraints. Incorporated cross-elasticity to prevent cannibalization between nearby properties. Deployed output dashboards on Tableau and internal BI tools with explainability layer.
Professional Experience
2+ years driving impact through data, analytics, and innovation
Operations Support Specialist
What I Did
Support the Admission Services team by managing enrollment-related workflows in Salesforce, ensuring accurate student records, and maintaining smooth processing during high-volume admission cycles. Handle Enrollment Cases, update student financial actions when needed (refunds, tuition credits, and charges), and assist with prospective student outreach through contact card entry. Collaborate closely with staff and student workers, provide peer training on Salesforce processes, and help maintain data accuracy and operational consistency across terms (Summer, Fall, Winter).
Key Achievements
Revenue Analyst
What I Did
Worked in the Central Revenue Management team, supporting daily pricing operations and ensuring the accuracy of OYO's automated pricing algorithm. I monitored occupancy and booking patterns across properties, validated algorithm-generated rates, and helped revenue managers make informed decisions by identifying pricing gaps, parity issues, and demand fluctuations.
Key Achievements
- •Analyzed how inputs like occupancy deltas, elasticity, base factor, floor price, ceiling price, and virtual floor adjustments influenced the system's final price recommendations
- •Identified cases where the algorithm mispriced properties and highlighted fixes such as threshold adjustments or override rules
- •Ensured algorithm outputs always adhered to the core logic (final price staying between the allowed floor and ceiling)
- •Reviewed occupancy trends, booking pace, cancellations, and demand dips to flag pricing issues early
- •Compared OYO rates against competitors to catch overpricing/underpricing and recommended corrections to regional teams
- •Coordinated price overrides when algorithm outputs didn't align with on-ground conditions
- •Analyzed how pricing changes impacted ARR, occupancy, and booking conversions
- •Supported the shift of demand across OTA, web/app, and walk-in channels by identifying channel-specific pricing issues and take-rate implications
- •Identified rate mismatches across OTAs (Booking.com, Agoda, etc.) and OYO's own channels
- •Ensured that incorrect parity wasn't feeding wrong signals back into the algorithm
- •Created templates for pricing audits, booking pace checks, and daily alerts
- •Cleaned and organized raw booking and revenue data for fast analysis and decision-making
- •Summarized key issues such as mispricing, demand anomalies, and realization drops
- •Provided clear, actionable recommendations aligned with algorithm logic and regional constraints
Associate
What I Did
Worked on analytical projects spanning investment pricing decisions and commodities forecasting. Built data-driven frameworks to support financial evaluation and trading strategy, combining feature engineering, modeling, and validation techniques tailored to business-specific requirements.
Key Achievements
Built a pricing and NPV evaluation framework to help the client assess the financial viability of future investments. Worked closely with business stakeholders to understand cost structures, payback assumptions, and scenario logic, and translated these into a data-driven pricing model.
- •Designed a Python-based NPV calculation workflow incorporating discount rates, payback periods, and demand assumptions
- •Engineered 20+ macroeconomic and operational features used for pricing simulations
- •Created standardized templates and scripts to quickly evaluate new deals using consistent assumptions
- •Automated repeat calculations and scenario tests, cutting manual effort significantly and reducing analysis errors
- •Delivered clear pricing recommendations backed by sensitivity analysis and demand scenarios
Developed an LSTM-based forecasting model for the Brent vs. Gas Oil spread using long-term commodity, macroeconomic, and futures data. Focused on signal generation, feature engineering, and temporal validation to improve prediction stability.
- •Built a multi-step LSTM forecasting pipeline for commodity spreads
- •Aggregated and cleaned long-history daily data from multiple sources (energy markets, macro indicators)
- •Engineered features such as lagged spreads, volatility bands, rolling statistics, and economic factors
- •Implemented walk-forward cross-validation to mimic real-world trading conditions
- •Created reusable modeling components adopted by teammates for similar time-series work
- •Provided periodic insights and analysis to help guide trading strategy reviews
Skills & Technologies
Tools and technologies I use to build amazing things
Programming Languages
Building solutions with modern languages
Machine Learning
Creating intelligent systems
Data Analytics
Transforming data into insights
Cloud & Tools
Deploying at scale
Specializations
Domain expertise areas
By the Numbers
Let's Connect
Open to opportunities in Data Science, ML Engineering, and Analytics roles
Ready to Chat?
Whether it's a project, job opportunity, or just a friendly hello, I'd love to hear from you!
Send me an email