KS
Karun Singampalli

Ask me anything.

I'm Karun's AI — I know his work, stack, and projects inside out.

Data Scientist & ML Engineer

Venkatakarun
Singampalli

Senior Data Scientist  ·  Production ML  ·  Agentic AI  ·  LLM Systems  ·  MLOps

Senior Data Scientist and ML Engineer with 4+ years of experience designing and shipping production ML systems — from hybrid recommendation engines and LLM-powered analytics pipelines to real-time fraud detection with automated drift monitoring. Based in Dallas, Texas.

Impact at a Glance
60 hrs
Weekly analyst hours eliminated via LLM pipeline
Banking · Symplore
26
RAG analytical workflows automated
GPT-4 · LangChain
20%
Accuracy gain over predecessor models
Fraud Detection
250+
Daily dashboard views
Power BI · Tableau
Core Expertise

What I Build

Six capability areas — each backed by production deployments

01
Agentic AI Systems

Multi-agent orchestration using LangGraph StateGraph — routing, retrieval, reasoning, and explainability agents in coordinated sequence.

LangGraphLangChainChromaDB
02
LLM & RAG Pipelines

Production RAG with hybrid retrieval, RAGAS hallucination evaluation, and prompt engineering for banking-scale accuracy.

GPT-4RAGRAGAS
03
Recommender Systems

Hybrid engines combining collaborative filtering (ALS, SVD, LightFM) with BERT content-based models and business rule layers.

ALSLightFMBERT
04
Real-Time ML Pipelines

Streaming inference with Kafka ingestion, sub-second scoring, and Prometheus/Grafana drift monitoring before accuracy degrades.

KafkaXGBoostPrometheus
05
MLOps & Deployment

Airflow DAGs, Docker containerization, REST API serving, MLflow versioning, and automated retraining at drift thresholds.

AirflowDockerMLflow
06
NLP & Text Analytics

Classification at 89% accuracy — TF-IDF, LSTM, BERT — deployed on 200K+ records with k-fold validation and SHAP explainability.

LSTMSHAPLIME
Completed Projects

Featured Work

Use the arrows to flip through project slides — GitHub, AI insights, and PDF for each

Agentic AI
Agentic Patron Intelligence System

Multi-agent LangGraph system — Retrieval, Analytics, Recommendation, and Explainer agents over ChromaDB.

⌥ GitHub
🤖
Slide 1 — Project Overview
Replace with your PPT screenshot
🏗️
Slide 2 — Architecture
🔍
Slide 3 — Retrieval Agent
📊
Slide 4 — Analytics Agent
🎯
Slide 5 — Recommendation Engine
💡
Slide 6 — SHAP Explainability
🐳
Slide 7 — Docker Deployment
📈
Slide 8 — Results & Metrics
🧪
Slide 9 — Evaluation
🚀
Slide 10 — Live Demo
1 / 10
RAG + Evaluation
Production RAG System

Hybrid retrieval (ChromaDB + BM25), RAGAS evaluation framework, hallucination rate monitoring across a 50-query test set.

⌥ GitHub
🔍
Slide 1 — RAG Overview
Replace with your PPT screenshot
📄
Slide 2 — Chunking Strategy
🧮
Slide 3 — Embeddings
🗄️
Slide 4 — ChromaDB Store
⚖️
Slide 5 — Hybrid Retrieval
🤖
Slide 6 — Generation
📏
Slide 7 — RAGAS Evaluation
📊
Slide 8 — Results
🐳
Slide 9 — Deployment
🚀
Slide 10 — Live Demo
1 / 10
Real-Time ML
Real-Time Fraud Detection Pipeline

Kafka ingestion, XGBoost + Isolation Forest ensemble scoring, Prometheus drift monitoring. One-command Docker Compose startup.

⌥ GitHub
Slide 1 — Pipeline Overview
Replace with your PPT screenshot
📡
Slide 2 — Kafka Architecture
🧬
Slide 3 — Feature Engineering
🌲
Slide 4 — XGBoost Model
🔭
Slide 5 — Isolation Forest
🤝
Slide 6 — Ensemble Fusion
📉
Slide 7 — Drift Monitoring
📊
Slide 8 — Grafana Dashboards
📈
Slide 9 — Model Performance
🚀
Slide 10 — Deployment
1 / 10
ML Platform
Smart Library ML Platform

Full ML lifecycle — Airflow ETL, Kafka ingestion, hybrid recommender (ALS + BERT + rules), FastAPI deployment, Prometheus monitoring.

⌥ GitHub
📚
Slide 1 — Platform Overview
Replace with your PPT screenshot
📥
Slide 2 — Data Ingestion
⚙️
Slide 3 — Airflow ETL
🧬
Slide 4 — Feature Engineering
🤝
Slide 5 — Collaborative Filtering
📝
Slide 6 — Content-Based NLP
Slide 7 — Hybrid Engine
📉
Slide 8 — Drift Monitoring
📊
Slide 9 — Dashboard
🚀
Slide 10 — Results
1 / 10
In Progress

Under Construction

Ideas currently being built — check back soon

🔬
LLM Fine-Tuning Lab
⚡ Building now

Fine-tuning Mistral-7B with LoRA adapters on domain-specific data. Comparing base vs fine-tuned performance on retrieval and generation tasks with automated evaluation.

Mistral-7BLoRAPEFTHuggingFace
🧪
A/B Testing Framework
⚡ Building now

Statistical A/B testing infrastructure comparing collaborative vs hybrid recommender models — measuring lift in Precision@10, Recall@10, and NDCG with significance testing.

SciPyStatsmodelsStreamlitPostgreSQL
Career History

Professional Experience

4+ years across ML engineering, data science, and product development — India & USA

Symplore Inc.
Jul 2025 – Present
Novi, MI · Remote
Senior Data Scientist — ML Engineering
Symplore Inc.
  • Engineered end-to-end ML and data engineering across 2 production platforms — Smart Library Analytics and TRIPGO — processing 8M+ records across 7+ data domains with 30% data-quality improvement
  • Built hybrid recommendation platform (ALS, SVD, BERT, rules) improving Precision@10 by 15% over baseline across 250K+ catalog items; inference at <300ms P95 latency supporting 1K+ requests/hour
  • Developed 12+ Airflow ETL workflows at 99%+ reliability; automated ML lifecycle with 10+ DAGs reducing manual operations by 50%; Prometheus/Grafana monitoring across 10+ pipelines
Intelligenie LLC
Sep 2024 – May 2025
Dallas, TX · Remote
Data Scientist
Intelligenie LLC
  • NLP classification on 200K+ Reddit posts — 89% accuracy, 25% improvement over baseline using TF-IDF, n-grams, and LSTM
  • 8 engineered features for stress detection; validated with k-fold cross-validation
360DigiTMG
Jan 2022 – Jul 2022
India · Remote
Data Analyst Trainee
360DigiTMG
  • Analyzed 3M+ sales records in SQL; built 5 Power BI dashboards replacing weekly manual reports
FAVO Robotics
Jan 2020 – Dec 2021
India · On-site
Product Development Analyst
FAVO Robotics
  • Analyzed 100K+ manufacturing and robotic data points using Python and SQL; built 15+ operational KPIs covering production output, cycle time, defect rates, and equipment utilization
  • Designed 5-axis robotic arm in SolidWorks with ANSYS FEA validation; engineered Python/ROS control workflows — improving motion accuracy by 55% and reducing manufacturing defects by 30%
M.S. Data Science
University of Texas at Arlington
Arlington, TX · 2022 – 2024
B.Tech Mechanical Engineering
Mahindra University
Hyderabad, India · 2016 – 2020
Contact

Let's Connect

Open to senior Data Scientist and ML Engineer roles at companies building serious ML infrastructure. Most effective on problems where production accuracy, pipeline reliability, and system design all matter equally.

Available for senior ML & Data Science roles · Dallas, Texas