// PLAYER_CARD — LordJordan02
Akhilesh Babu
Tumati
Full-Stack & AI Engineer
Alexandria, VA
_
Dossier
Full-stack and AI engineer with experience spanning enterprise (Honeywell, UPS), university research and platform work (Virginia Tech), and product-focused builds (AI try-on kiosks, fraud detection). Comfortable across the stack — Java/Spring Boot, TypeScript, React, Next.js, Python on the front, AWS, Kubernetes, Jenkins, MongoDB, PostgreSQL, Couchbase on the back — and equally at home shipping CI/CD pipelines, hardening security, or integrating generative AI. Treats Claude Code, MCP servers, and agent harnesses as everyday engineering tools.
Skills Arsenal
Languages
Frameworks
Cloud / DevOps
AI / ML
Databases
Mission Log
Full Stack Developer (Part-Time)
Virginia Tech- Built a scalable academic content platform with Next.js 15 and Express.js — cut first-contentful-paint from 2.4s to 1.4s (40% faster) and enabled sub-second search across 10+ research domains.
- Designed a fully serverless AWS stack (S3, CloudFront, Lambda) via CloudFormation, pairing it with MongoDB Atlas to slash operational costs by 25% over a comparable EC2 setup.
- Tuned Lambda cold-start latency from 980ms down to 540ms (45%) through connection pooling and backend optimization, keeping the platform snappy even during peak ingestion windows.
- Locked down access with MFA, a 5-tier RBAC system, and JWT token blacklisting — bringing the platform into full compliance with university data governance standards.
Full-Stack AI Engineer — Sponsored Capstone
Marine Corps Community Services (MCCS)- Built a real-time ML fraud detection pipeline in Python (XGBoost) for the MCCS e-commerce platform — used GAN-synthesized data to fix class imbalance and score 15,000+ daily transactions in under a second.
- Hit 95% accuracy and 98% recall after tuning and head-to-head comparison of multiple classifiers, catching suspicious transactions earlier and with fewer false alarms.
- Engineered a RAG assistant (Gemini, LangChain, FAISS) to auto-generate Explainable AI audit reports grounded in real fraud context — replaced ~12 hours/week of manual PDF prep with 150+ reports a month.
- Shipped a live monitoring dashboard (React, Node.js, Recharts) to surface fraud patterns and streamline model retraining, wired directly into the existing e-commerce infrastructure.
Software Engineer
UPS- Drove 20+ monthly zero-downtime releases for high-traffic logistics microservices through Jenkins CI/CD into OpenShift/Kubernetes, cutting deployment errors by 30% across UPS's global delivery network.
- Kept distributed Java services at 99.99% availability by wiring Grafana and Dynatrace alert rules to Couchbase NoSQL cluster health — catching and resolving P1 incidents before they hit customers.
- Owned schema design and query optimization for production Couchbase databases handling 10M+ daily address lookups, shaving average query latency by 35%.
- Executed global data migrations across APAC/EU/US using Red Hat AMQ Streams (Kafka) and MirrorMaker 2, ensuring synchronized production data and automated backups to Azure for disaster recovery.
Software Developer Engineer
Honeywell- Shipped full-stack features on ASDS — a B2B e-commerce platform serving 500+ enterprise aviation clients — building 15+ reusable Angular components backed by Java Spring Boot + JPA ORM, cutting frontend dev time by 25%.
- Wrote 200+ end-to-end test cases for payment gateway and revenue-critical flows using Java Selenium and Katalon Studio, achieving 85% regression coverage and reducing QA cycle time by 40%.
- Embedded Coverity and Blackduck security scans into every CI/CD build via Bamboo and Bitbucket, cutting security vulnerabilities by 60% and ensuring zero critical-CVE deployments.
Projects
PROJECT — 01
MCCS Fraud Detection Platform
VT Integrated Product Design — MCCS Sponsor
Real-time inference pipeline in Python using XGBoost trained on GAN-augmented synthetic fraud samples, achieving sub-second risk classification across 15,000+ daily transactions. Includes a RAG-powered Gemini chat assistant over fraud cases and React/Tremor dashboards.
Python
XGBoostGeminiMongoDBReactPROJECT — 02
Evol Jewels — Virtual Try-On Kiosk
AI retail kiosk · Fountane Hackathon (2nd place)
High-impact retail kiosk built with Next.js 15 and TypeScript, integrating fal.ai generative workflows to improve virtual try-on image quality by 30%. Catalog managed via Prisma ORM and PostgreSQL with QR-code lead capture.
Next.jsTypeScript
PrismaPostgreSQLPROJECT — 03
CS Ethics Archive
Virginia Tech — academic repository
Scalable academic repository with sub-second search across 10+ domains, built on Next.js and Express with serverless AWS infrastructure and MongoDB Atlas. Hardened with MFA, 5-tier RBAC, and JWT token blacklisting.
Next.jsExpress
AWSMongoDBResearch
IEEE ICAECT 2023
Tumati, A. B., Gangaraju, R., Mannepalli, B. R., & Alluri, B. K. (2023, January). Face Invariant Classification and Detection of Mythology Characters Using Custom Dataset (ClaDeMuC-CD). In 2023 Third International Conference on Advances in Electrical, Computing, Communication and Sustainable Technologies (ICAECT) (pp. 1-8). IEEE.
IJHS 2022
Natarajan, K., Gangaraju, R., & Tumati, A. B. (2022). Hybrid ML and DL models for flood level prediction. International Journal of Health Sciences.
Education
Virginia Tech
Master of Engineering, Computer Science
Vellore Institute of Technology
B.Tech, Computer Science