Post Job Free
Sign in

Conversational AI & NLP Data Scientist

Location:
Houston, TX
Posted:
November 02, 2026

Contact this candidate

Resume:

Likhil Yadav Arala

***** **** *** **, *******, TX 77084

217-***-**** ************@*****.***

Summary

Data Scientist with 4+ years building NLP and machine learning systems across healthcare, retail, and logistics. Hands-on with Transformer models, embeddings, text classification, and LLM fine-tuning, plus conver- sational AI design covering intent and entity modeling. Comfortable taking models from feature engineering through production deployment, latency tuning, and evaluation with golden datasets. Experience

Accenture May. 2025 – Present

Data Scientist, Conversational AI & NLP United States

•Designed and shipped a conversational AI assistant for logistics dispatch, covering conversation design, intent and entity modeling, and error analysis that cut misrouted requests by 27% over the first quarter.

•Built text classification and embedding pipelines in Python using Transformer models and LLMs, fine-tuning on 40K+ labeled support transcripts to improve routing accuracy across five request categories.

•Partnered with operations leads to define a golden dataset and LLM-as-a-Judge evaluation framework, giving the team a repeatable way to score assistant responses before each production release.

•Applied feature engineering and time-series forecasting to telematics data, blending weather, terrain, and payload signals into models that flagged at-risk delivery segments ahead of dispatch.

•Reduced inference latency by 35% through KV caching, batching, and prompt trimming, letting the assistant scale to 2,000+ daily queries on existing AWS infrastructure without added compute spend.

•Mentored two analysts on model evaluation and prompt design, and presented assistant performance metrics to senior leadership, which secured budget to extend the system to two additional depots. Cognizant Jun. 2022 – Dec. 2023

Machine Learning Analyst India

•Supported a top-10 pharmaceutical client’s patient engagement program, where the core challenge was predicting churn early enough for outreach teams to act before patients disengaged from therapy.

•Built text classification models over clinical notes and engagement signals, combining embeddings with XGBoost to flag 60-day attrition risk and feed a prioritized outreach queue.

•Engineered time-series forecasting pipelines in Python (statsmodels, scikit-learn) and SQL/Snowflake, running multi- TB data preprocessing jobs on Azure Synapse Spark.

•Automated forecast delivery through Snowflake Tasks and embedded data-quality monitors in Azure Data Factory, so downstream model evaluation ran on clean inputs without manual checks.

•The retention program rescued 28% of at-risk patients, and the forecasting suite improved MAPE by 92%, giving the client a reliable weekly demand signal for the first time.

•Mentored three analysts on machine learning best practices and cloud orchestration, and translated model findings into roadmaps that stakeholders could act on without a data background. Express Stores Jun. 2021 – Jul. 2022

Junior Data Analyst India

•Drove omni-channel analytics unifying in-store POS and e-commerce feeds to optimize product mix, pricing, and customer engagement across 1,000+ SKUs.

•Partnered with merchandising to launch a new web storefront, embedding real-time recommendation logic that surfaced relevant products based on browsing and purchase history.

•Built data preprocessing pipelines in Python, PySpark, and T-SQL/Snowflake, with CI/CD through GitLab and Docker to keep releases consistent.

•Authored 10+ interactive dashboards in Power BI and Tableau, and automated scraping jobs with Airflow that saved 30+ developer hours per pricing cycle.

•Insight-driven assortment and promotion strategies lifted store sales 22%, while a governance framework raised reporting accuracy to 98% across teams.

Technical Skills

NLP & LLMs: Transformer Architecture, Large Language Models (LLMs), Text Classification, Embed- dings, Intent & Entity Modeling, Conversation Design, Model Fine-Tuning, LLM-as-a- Judge, Golden Datasets, Agentic AI, Multi-Agent Architectures ML & Analytics: Python (Pandas, NumPy, SciPy, scikit-learn), XGBoost, statsmodels, ARIMA/SARIMA, Feature Engineering, Time-Series Forecasting, Data Preprocessing, Model Evaluation, A/B Testing

Production & MLOps: Inference Optimization, KV Caching, Latency Reduction, Scalability, Model Deployment, MLflow, Docker, Kubernetes, Apache Airflow, Git/GitHub/GitLab CI/CD Cloud & Data: AWS (SageMaker, S3, Lambda), Azure Synapse, Azure Data Factory, Databricks, Snowflake, SQL (PostgreSQL, T-SQL)

Visualisation & BI: Power BI, Tableau, Plotly Dash, Matplotlib, Jupyter Notebooks Leadership & Collaboration:Mentoring, Cross-functional coordination, Stakeholder presentations, Data governance Projects

Conversational AI Assistant with LLM Evaluation Python, Transformers, LangChain, FastAPI, Docker, AWS

•Fine-tuned a Transformer-based LLM on 25K support conversations for intent classification and entity extraction, reaching 89% intent accuracy across 12 categories.

•Designed a multi-agent pipeline with a router agent and specialist agents, evaluated against a curated golden dataset using LLM-as-a-Judge scoring for relevance and groundedness.

•Cut response latency from 2.4s to 1.1s using KV caching and batching, then deployed via FastAPI and Docker on AWS with monitoring for drift and fallback handling. Education

Eastern Illinois University Charleston, Illinois

Master of Science, Computer Technology 01/2024 – 12/2025



Contact this candidate