SUKESH S

Building on the Edge of AI & Software Engineering

B.Tech in Artificial Intelligence & Data Science @ SNU Chennai. AI Fullstack Developer & ML Engineer — production ML pipelines, LLM-orchestrated reasoning systems, and full-stack platforms.

GITHUB ACCESS
GPA: 8.45

THE MANIFESTO

I thrive at the intersection of Deep Intelligence and Scalable Architecture. I don't just train models; I build the infrastructure that allows them to serve reality.

CURRICULUM SPECIALIZATION

Data StructuresOperating Systems Analysis Of AlgorithmsArtificial Intelligence Machine LearningDatabases

POWER STACK

PyTorchTensorFlow scikit-learnXGBoost LangChainNeo4j SHAPFastAPI Next.jsPostgreSQL MongoDBDocker KubernetesRedis

BASE OF OPERATIONS

Shiv Nadar University Chennai

Bachelor of Technology - AI & Data Science (2024 - 2028)

ENCRYPTED COORDS

EMAIL: ssukesh2007@gmail.com

PHONE: +91 8807000282

LOC: Chennai, India

Combat Logs

Strativa

AI Fullstack Developer / ML Engineer

MAY 2026 - JUNE 2026
REMOTE
  • Fraud Intelligence: Built a Fraud Investigation Intelligence Platform using XGBoost, SHAP, Neo4j, and graph community detection to identify fraud rings and suspicious account networks across large financial datasets.
  • Alert Compression: Engineered an alert-compression pipeline grouping related fraud alerts into prioritized investigation clusters via network-level analysis, cutting analyst workload significantly.
  • Model + Explainability: Trained and tuned an XGBoost fraud classifier with feature engineering, class-imbalance handling, and k-fold CV, validated via precision-recall/ROC-AUC, with SHAP-based per-decision explainability for analyst trust.

Shiv Nadar University – ML Lab

Machine Learning Research Intern

APRIL 2026 - PRESENT
CHENNAI, INDIA
  • State Estimation: Implemented and compared SLAM, Kalman Filter, Extended Kalman Filter, and Adaptive Monte Carlo Localization on a toy robot platform to study state estimation for indoor navigation.

Kazebuilds

Full Stack Developer

MARCH 2026 - JUNE 2026
REMOTE
  • Full-Stack Delivery: Architected and delivered a production-grade e-commerce platform for an energy-sector client end-to-end, owning the full SDLC from UI/UX through backend architecture to infrastructure deployment.
  • Scalable Infrastructure: Built a server-rendered Next.js + TypeScript frontend backed by a RESTful FastAPI backend and normalized PostgreSQL schema; containerized with Docker and configured Nginx as reverse proxy for production deployment.

Saint Louis University

Data Analysis Intern

NOV 2025 - DEC 2025
REMOTE
  • Campaign Analytics: Analyzed outreach and engagement datasets using Python and Pandas to identify high-engagement audience segments and actionable insights.
  • Visualization Intelligence: Designed Power BI and Tableau dashboards that directly informed budget reallocation decisions.

Mission Artifacts

MISSION: KAZEPOWER

Real-world, production e-commerce platform built end-to-end for an energy-sector client — from UI/UX through backend architecture to deployment. Server-rendered Next.js frontend, RESTful FastAPI backend, normalized PostgreSQL schema for catalog/cart/orders, secure checkout, Dockerized stack behind Nginx.

Next.jsTypeScriptFastAPI PostgreSQLDockerNginx

MISSION: KEYSER INTELLIGENCE

LLM-orchestrated competitive intelligence engine that ingests omnichannel market signals (web, social, job boards, pricing, news) and reasons over them via chained LLM inference. Engineered an N-dimensional whitespace computation engine, per-tenant schema-level data isolation, async multi-source workers, and a streaming AI copilot for strategic querying.

LangChainFastAPICelery PostgreSQLRedisReact

MISSION: AUTOSPHERE

Real-time vehicle telemetry platform ingesting high-frequency streams via WebSockets, running Isolation Forest-based anomaly detection, and triggering automated service alerts. MongoDB time-series data models with FastAPI microservices for vehicle registry and workflow automation.

PythonFastAPIMongoDB Redisscikit-learnWebSockets

MISSION: EDUCREATE

Vision-grounded RAG tutoring platform: users upload images and get concept explanations, auto-generated questions, and chatbot support. Fine-tuned ViT-B/16 for image-to-concept classification (88.4% top-1 accuracy across 34 STEM categories) plus a FAISS-backed RAG chatbot (MRR 0.79, ~2.1s/image question latency).

ViT-B/16FAISS LangChainReactMongoDB

MISSION: ANOMNET

Deep learning system predicting Remaining Useful Life for turbofan engines from multivariate sensor time series. A 2-layer LSTM over 30-cycle sliding windows models degradation dynamics on NASA's CMAPSS (FD001) dataset, with RUL capped at 125 cycles for training stability.

PyTorchLSTM PandasNumPyscikit-learn

TROPHY ROOM

KHacks 3.0 WINNER

Anna University // Web3 Track

Secured first place in the highly competitive Web3 track at the national level hackathon.

GDG Hackathon WINNER

Google Developer Groups // SNUC

Overall winners of the GDG-hosted hackathon for innovative AI implementation.

Reliance Scholar

Top 5000 in India

Recipient of the prestigious merit scholarship for academic and technical excellence.

SNUCHacks Runners

Project: Vision Transformers Integration

EY Techathon Finalist

Automotive Digital Twin Innovation

VOID Hackathon Top 10

Healthtech Category Finalists

Top 10 Finisher - Glytch (National Level), VIT Chennai