Sinchana G V
Turning data into decisions that matter
2+ years building AI systems, analytics pipelines, and data-driven products. MS Data Science, University of Maryland (GPA 3.81). MCP AI Hackathon winner. Streamlit Creator. Authorized to work in the US.
Invoice Agent: AI-Powered X12 Processing
End-to-end AI workflow parsing, validating, and enriching EDI X12 invoices using RedisVL vector search, semantic matching, and LLMs.
FinTech Risk Analytics
dbt + BigQuery platform analyzing $152M in customer exposure with ARIMA+ forecasting.
Customer Churn on AWS
Ensemble ML pipeline predicting churn 30 days in advance. SageMaker + SHAP feature analysis.
Poetic Device Classifier
Fine-tuned Mistral 7B on 13,900 stanzas using LoRA. 20% precision improvement.
Multimodal Drift Detector
Statistical drift detection with LLM-powered root-cause explanations and Slack alerts.
Engineer by training,
builder by instinct
Currently at Connyct Inc. as an AI and Data Science Technology Engineer, I engineer production AI workflows that automate enterprise processes. My invoice processing system (built with RedisVL and LLMs) earned 3rd place at the MCP AI Hackathon, reducing manual review time by 60%.
At Mann+Hummel, I deployed 10+ interactive dashboards improving decision-making by 20% and automated ETL saving 15+ hours weekly. At Accenture, I reduced Nissan EU's job failures from 40% to under 10%, saving an estimated €2M+ annually.
I serve as a Graduate Teaching Assistant for ML at UMD supporting 100+ students, and I'm a proud member of the Streamlit Creators Program.
What I work with
Where I've made an impact
New York, NY (Remote)
- Designed and deployed AI workflows integrating LLMs into enterprise applications, enhancing automation across multiple business units
- Built scalable data pipelines using Python, Redis, and AWS supporting real-time analytics processing thousands of transactions daily
- Reduced incident response time by 60% through automated reporting workflows surfacing data quality insights and anomalies
- Developed Streamlit dashboards for real-time performance monitoring for Sales and Customer Success teams
College Park, MD
- Assisted professors teaching Machine Learning and Data Science to 100+ graduate students
- Conducted office hours, graded assignments, and provided feedback on Python, ML algorithms, and statistical analysis
- Developed supplementary code examples to enhance student understanding of complex ML concepts
Raleigh, NC
- Designed and deployed 10+ interactive dashboards using Power BI, Tableau, and Python, improving decision-making by 20%
- Automated ETL workflows reducing manual processing by 15 hours/week and improving data accuracy by 30%
- Applied clustering and regression to forecast inventory, reducing stockout incidents by 18%
- Performed EDA on 500K+ records to surface data storytelling narratives for leadership
Bangalore, India
- Processed and transformed registration data from 10,000+ VINs daily for Nissan EU using Python and SQL
- Reduced job failures from 40% to under 10%, saving an estimated €2M+ annually
- Built reporting workflows improving system reliability from 60% to over 90%
- Partnered with cross-functional teams to translate business questions into technical implementations
Production-grade systems,
measurable impact
Invoice Agent: AI-Powered X12 Processing
End-to-end AI workflow parsing, validating, and enriching EDI X12 invoices using RedisVL vector search and semantic matching. Streamlit dashboard for real-time anomaly monitoring.
FinTech Risk Analytics Dashboard
Risk analysis platform for 2,965 customers with $152M exposure. ARIMA+ forecasting with BigQuery ML. 32% high-risk customers hold 3x disproportionate exposure.
Multimodal Data Drift Detector
Drift detection combining KS test, PSI, and z-score with multimodal context from PDFs, screenshots, and logs. AI root-cause explanations with Slack alerts.
Customer Churn Prediction on AWS
ML pipeline predicting churn 30 days ahead using ensemble methods. SageMaker deployment with QuickSight monitoring and SHAP-based feature importance.
Poetic Device Classifier (LLM Fine-tuning)
Fine-tuned Mistral 7B on 13,900 annotated stanzas using LoRA. FAISS retrieval + GPT explanations. Processes 1,000+ stanzas per minute.
DC Metro Ridership Analytics
3+ years of WMATA data across 98 stations. Seasonal decomposition, peak analysis, anomaly detection near venues, with dynamic scheduling recommendations.
COVID-19 Case Fatality Rate Analysis
Comprehensive statistical analysis of global COVID-19 CFR data. EDA and wrangling on large-scale pandemic datasets. ggplot2 visualizations with R Markdown report covering statistical interpretations and actionable takeaways.
Awards & honors
3rd Place, Redis VL Innovator Category
MCP AI Hackathon · AI-powered invoice processing agent using RedisVL vector search, semantic matching, and real-time anomaly detection.
2025Apex Award Winner
Accenture, India · Outstanding performance and measurable business impact. Honored for improving system reliability from 60% to 90%+.
May 2023Client Value Creation Award
Accenture, India · Identifying root causes of registration failures and reducing operational costs by an estimated €2M+ annually for Nissan EU.
Dec 2022Streamlit Creators Program
Selected as a Streamlit Creator for contributions to the open-source data app ecosystem and high-quality Streamlit application development.
2025 – PresentLet's build
something great
Actively seeking full-time roles in Data Analytics, Data Science, Business Analysis, Product Analytics, and Risk Analytics. Whether you want to discuss a role or explore collaboration, I'd love to connect.