Currently at Truist Bank · Charlotte, NC

Sakshi
Doshi

Data Scientist

AI Engineer

Risk Intelligence Builder

LLM Systems Architect

I build AI systems that turn raw data into executive-level intelligence — from automated risk briefings for Chief Risk Officers to fraud detection across 70 billion transactions.

View My Work Get In Touch

SCROLL

Hours Saved / Year

Hours via Sentinel100

Complaints Flagged

KRIs Tracked

Transactions Processed

Fraud Recall

Career

Where I've Built Things

From detecting fraud at fintech scale to automating intelligence for a top-15 US bank's Chief Risk Officer.

Data Scientist II

Truist Bank

📍 Charlotte, NC

Sept 2025 – Present

AI-Powered Risk Innovation & Automation

Deployed an AI workflow on Rogo.ai monitoring 13 key risk domains — daily automated briefing to the Chief Risk Officer and 14 direct reports, saving an estimated 2,000+ manual hours annually with built-in AI-powered QA.

Developed Sentinel100 with Rogo.ai + Python: weekly adverse news & market surveillance across Truist's top 100 clients and 5 risk signals, saving 250+ manual hours/year by replacing Excel-based filtering.

Built a Python NLP Pipeline scoring and flagging 213K+ Tier-1 complaints using weighted keyword detection and TF-IDF/KMeans clustering with GitLab Duo AI assistance.

Led automated monthly reporting for 891 Key Risk Indicators — saving 70+ hours/month + 60 hours/quarter, delivered to the CRO, Risk Committee, Board, and all Business Units.

Data Scientist I

DataSeers.ai

📍 Alpharetta, GA

July 2023 – Sept 2025

Built fraud detection models for Human Sex Trafficking & Elderly Financial Abuse using XGBoost + DBSCAN — 85% recall, 70% precision, uncovering fraud rings of 1,500+ accounts via graph-based analysis.

Developed a RAG chatbot (React, Node, Kafka, Flask, Llama3) for ECL developers — reduced language-related support queries by 55%.

Cleaned and classified 90% of unclean merchant names in bank transactions using Llama2, BERT Transformers, and Regex, improving downstream alert quality.

Developed risk ratings (Address, Account, Identity, Account Holder) using Logistic Regression + XGBoost across 70 billion transactions — improved fraud prevention and compliance monitoring by 60%.

Software Engineer Intern

MathWorks

📍 Natick, MA

May 2022 – Aug 2022

Enhanced readability of the Allocation Matrix by over 90% by converting it into Allocation Diagrams using MATLAB, C++, and JavaScript. Developed a comprehensive Dataflow for the Allocation Diagram Architecture using the Simulink Codebase.

Independently reflected 95%+ of the Matrix as an Allocation Diagram, completing the full feature end-to-end and addressing various logical design problems.

Contributed to the Architecture Modeling Web Development Project as part of an 11-person System Core Architecture Team.

Web Developer, Graphic Designer & IT Support

Center for Women & Community

📍 Amherst, MA

Feb 2022 – May 2023

Provided web development, graphic design, IT support, and community outreach services during graduate studies at UMass Amherst.

Software Engineer & Full Stack Intern

Centre for Development of Advanced Computing (C-DAC)

📍 Pune, India

Aug 2020 – Apr 2021

Mesh Generation for Flood Simulation: Reduced mesh triangles by 27% without sacrificing accuracy by replacing irregular triangular meshes with regular ones during simulation generation.

Optimized the software development process using Param Shakti Supercomputer by 40% through parallel computation.

Built a Flood Dissemination Web Portal that increased accessibility by ~98% by depicting real-time flood simulations from triangle-predicted data.

Improved portal functionality by 85% using React, MongoDB, Express, Node.js, and libraries including Leaflet.js, Geotiff.js, and Chart.js.

Machine Learning Engineer Intern

Securly

📍 Pune, India

Aug 2020 – Dec 2020

Eliminated 90% of redundant email content — signatures and disclaimers — using NLP and Regex data cleaning techniques.

Built a system to analyze student emails and detect early signs of mental health conditions related to self-harm and suicide risk.

Analyzed more than 5 billion activities — the system has been credited with saving 1,937 lives.

Research & Projects

Built to Win

National hackathon recognition, Kaggle benchmark-beating results, and first-place competition finishes.

01 / SMART INDIA HACKATHON · BPRD

Deepfake Detection using Deep Learning

🏆 Led team to nationals · 96% accuracy

Classified deepfakes using EfficientNet B7 for the Government of India's BPRD agency. Built a cross-platform Flutter mobile app and a full MERN web portal deployed on Microsoft Azure.

EfficientNet B7FlutterMERNAzureDeep Learning

02 / MEDICAL AI · KAGGLE

Pulmonary Fibrosis Progression Prediction

🥇 Beat Kaggle winner · LLL: −6.8303

Outperformed the Kaggle competition winner predicting Forced Vital Capacity progression in pulmonary fibrosis patients using ResNet + Multiple Quantile Regression.

ResNetMedical AIQuantile RegressionKaggle

03 / COURSE COMPETITION

Pawpularity Contest — Pet Photo Popularity

🥇 First rank in course · RMSE: 17.93

Forecasted pet photo popularity by training 56 models — SWIN Transformer, Transfer Learning, CNNs, and ensemble methods with extensive feature engineering.

SWIN TransformerTransfer LearningCNNsEnsemble56 Models

Giving Back

AI for Good

Volunteer Project

Clinical Trials Finder

Cerebral Palsy Research Network

Built an Agentic AI serverless application on Azure that translates ClinicalTrials.gov eligibility criteria into plain language — making clinical trials accessible to patients and caregivers who aren't medical professionals. Auto-deploys via GitHub Actions CI/CD with $0 idle cost.

Agentic AI Azure Container Apps GPT-4o Microsoft Foundry GitHub Actions

🤖

AI Engine

Azure OpenAI via Microsoft Foundry + GPT-4o

☁️

Infrastructure

Azure Container Apps · Serverless · $0 idle

🚀

Deployment

Auto-deploys via GitHub Actions CI/CD

💙

Impact

Plain-language trial eligibility for CP patients

Technical Arsenal

Skills & Tools

Full-stack AI engineering — from raw data pipelines to production LLM deployments at enterprise scale.

Languages

PythonRC#JavaScriptHTMLCSS

Frameworks

PyTorchFlaskKafkaPandasNumPyScikit-learnLangChainStreamlitReactNode.js

AI & LLMs

Agentic AIMulti-AgentRAGRogo.aiLlama2/3BERTGPT-4oTF-IDFKMeans

Databases

MongoDBSQLRedisMySQLElastic

Cloud & DevOps

AWSMicrosoft AzureGoogle CloudContainer AppsMS FoundryGitHub ActionsGitLab Duo

Academic Foundation

Education

🎓

Master of Science in Computer Science

Specialization in Data Science

University of Massachusetts Amherst

📅 May 2023📍 Amherst, MA

GPA: 3.86 / 4.0

🏛️

Bachelor of Engineering in Computer Engineering

Savitribai Phule Pune University

📅 April 2021📍 Pune, India

GPA: 8.6 / 10

Get In Touch

Let's Build Something Together

Open to interesting problems in AI, risk intelligence, and data systems.

sakshidoshi3@gmail.com LinkedIn GitHub