Sashi Kiran MaddineniAI, ML & Data Science

I build intelligent knowledge graphs, advanced LLM architectures, and predictive models that solve real-world problems.

View Projects → Contact Me

analytics-dashboard.app

Model Serving Analytics ● Live

Monthly throughput — records processed (M)

99.7%

Pipeline uptime

2.4M

Events / day

12ms

Avg latency

AI Insights

Alert

Inference latency spiked 18% in Cloud Run deployment. Review FastAPI serving container.

2m ago

Insight

RAG query understanding accuracy improved to 92.4% after latest knowledge graph enhancement.

15m ago

Complete

Model deployment to Kubernetes cluster finished. All 47 tests passing.

1h ago

Suggestion

Consider optimizing the PyTorch tensor memory allocation — inference time would drop ~40%.

3h ago

What I solve

Your data isn't failing at scale.
It fails before scale in the foundation

Most AI problems aren't about the models. They're structural — disconnected data, fragile pipelines, and hallucinations nobody trusts.

🔀

Disconnected knowledge, poor entity resolution

Data lives in silos without semantic understanding. AI models struggle to provide accurate answers because the underlying context is fragmented.

CSV →

API →

DB →

✓

Unified

🧩

Fragile ML infrastructure that breaks at scale

Every deployment feels like a gamble. Models are pushed to production with no monitoring, no scaling strategy, and unpredictable latencies.

● 2 failures / week

→ 0 with monitoring

📉

AI tools that look smart but hallucinate

Models generate confident but incorrect answers. Users nod politely then abandon the tool because it lacks domain-specific guardrails.

Vanity metrics ✗

Actionable insights ✓

Decision framework ✓

Expertise

Built across the full data stack

From ingestion to insight — I own the pipeline end to end.

Machine Learning & AI

Expert

Designing LLM architectures, Knowledge Graphs, advanced RAG systems, and deploying predictive models at scale.

PyTorch RAG LangGraph LLMs NLP Databases

Data Science & Engineering

Expert

Building fault-tolerant ETL pipelines, scalable data lakes, exploratory data analysis, and statistical modeling for ML-driven insights.

Python SQL Databricks PySpark Airflow Azure DataFabric

Cloud & MLOps

Advanced

Cloud-native AI platforms, containerized model serving, scalable inferences, and cost-optimized orchestration.

GCP AWS Azure Kubernetes FastAPI

Analytics & Visualization

Advanced

Exploratory data analysis, statistical modeling, graph-based analytics, and building robust stakeholder dashboards.

Power BI Pandas Statistical Analysis Tableau

Selected Work

Projects that moved the needle

Real outcomes from real data problems.

Agentic AI 2025

ALFA – Agentic AI Agent

Built an AI Agent with LangGraph using Copilot to secure data, enable NL queries, integrate REST APIs, and enforce RBAC/guardrails.

LangGraphCopilotREST APIsPython

100% Secure via RBAC

LLMs & RAG 2025

NLP / Knowledge Graphs

Built an NLP system with LLM architecture and advanced RAG integrated with knowledge graphs, achieving 92% query understanding accuracy.

PythonLLMsRAGKnowledge GraphsSQL

92%

Query accuracy

Predictive Modeling 2024

Cryptocurrency Market Analysis Engine

Modeled market relationships as graph structures with Gini and Kolkata indices, improving cryptocurrency price prediction accuracy by 25%.

PythonGraph TheoryStatistical ModelingPandas

+25%

Prediction accuracy

Data Engineering 2024

End-to-End Data Management Pipeline

Designed a pipeline processing 100GB+ healthcare claims data using Hadoop/Spark/Hive, cutting payment collection time & achieving 95% accuracy.

HadoopSparkHiveAzure

-30%

Processing time

Career

Where I've built and shipped

Feb 2025 — Jan 2026

AI/ML Engineer

Vosyn

Developed entity-level connections between LLM outputs to enrich knowledge graph structures, trained models on GCP with PyTorch, and deployed pipelines on Kubernetes Cloud Run integrated with GCS. Optimized serving with FastAPI, improving latency by 30%.

Mar 2024 — Jun 2024

AI Engineer Intern

Tublian

Fine-tuned Llama3 chatbot with entity recognition and semantic matching for knowledge-based query resolution. Designed scalable ML models with 90% deployment success and built an AI-powered SQL agent using Copilot.

Mar 2021 — Dec 2022

Programmer Analyst / Data Scientist

Cognizant Technology & Solutions

Built ETL pipelines with Python/Spark (98.7% accuracy), developed predictive models improving KPI forecasting by 20%, optimized real-time processing with PySpark & Kafka, and delivered insights via Power BI.

Years in AI/Data

Developing predictive models and scalable ML architectures.

15+

ML Pipelines Shipped

Models serving predictions in production via robust ETL systems.

Certifications

AWS, Oracle Cloud, IBM, and CAP certified analytical professional.

How I Work

Built for reliable, production-grade delivery

Every project follows principles that keep systems running and stakeholders confident.

🧪

Tested & Monitored

Every pipeline ships with automated tests, data quality checks, and observability from day one.

Great Expectations / dbt tests
Pipeline SLAs and alerting
Automated data freshness checks

📐

Documented & Reproducible

Clean code, version control, and documentation that lets anyone pick up where I left off.

Git-based workflows
Containerized environments
Self-documenting pipelines

🤝

Collaborative & Transparent

I translate between technical and business — keeping stakeholders in the loop without the jargon.

Async-first communication
Weekly progress artifacts
Clear scope and timeline docs

Questions & Answers

Frequently asked

I work across the full AI/ML lifecycle — from building data pipelines and knowledge graphs to creating predictive models and reliable LLM applications. I specialize in turning fragmented data into robust intelligence.

Python, SQL, PyTorch, TensorFlow, LangChain, Kafka, Cloud Providers (AWS/GCP), and FastAPI. I pick the right tool for the job to ensure scalable and reliable ML production systems.

Both. I've built AI infrastructure for open-source AI projects, and I've delivered complex ML pipelines inside large enterprise environments with heavy governance requirements.

Absolutely. I integrate into existing workflows, adopt your team's coding standards, and document everything so the work lives on after my engagement ends. No vendor lock-in to my methods.

We start with a discovery call to understand your data landscape and goals. From there, I scope the work, define milestones, and deliver iteratively — usually with weekly check-ins and async updates.