Fig. 1. Architectural Schematic
Jayaditya Dev avatar in light modeJayaditya Dev avatar in dark mode
❯
AVAILABLE FOR FULL-TIME ROLES

Jayaditya Dev

Backend engineer. AI systems. Production-first.

Backend-focused engineer building reliable, production-oriented, and secure backend systems, APIs, deployment workflows, and service-oriented architectures. Currently building RAG pipelines and multi-agent systems at 7HiddenLayers.

Off-duty:🎸Guitar⚡Breaking Systems🏓Table Tennis🎯Valorant
GitHubLinkedInEmailDownload CV
Bengaluru, IN
Python (uv)◆FastAPI◆PostgreSQL◆PyTorch◆RAG Systems◆WebSockets◆Docker◆AWS Lambda◆SQLAlchemy◆Multi-Agent◆Computer Networks◆Linux◆Vector DBs◆Python (uv)◆FastAPI◆PostgreSQL◆PyTorch◆RAG Systems◆WebSockets◆Docker◆AWS Lambda◆SQLAlchemy◆Multi-Agent◆Computer Networks◆Linux◆Vector DBs◆
01 // PRODUCTION EXPERIENCE

AI Backend Infrastructure & Retrieval Systems

Direct engineering impact building production RAG pipelines and multi-agent systems at 7HiddenLayers.

7HiddenLayers

Internship

AI Backend Engineer Intern

June 2026 - Present
Bengaluru, IN

Designed incremental ingestion logic for a RAG-based legal document parsing system, ensuring partial document updates re-process only modified segments instead of the entire knowledge base.

Built client-facing query infrastructure where small language models retrieve from a centralized knowledge base, strictly constraining responses to verified source citations to eliminate hallucinations.

Constructed a multi-agent orchestration pipeline that automatically benchmarks creator posts against organizational SOPs, surfacing actionable revisions prior to approver review.

Core Stack:PythonuvFastAPIRAG SystemsSLMsMulti-Agent OrchestrationVector DBs
02 // ARCHITECTURE & IMPLEMENTATIONS

Production Systems & Flagship Repositories

Compact 2×2 engineering matrix spanning real-time backends, ML inference pipelines, and quantitative decision systems.

OctNovDecJanFebMarAprMayJunJulAugSep
Fig. 2.1,480 contributions, 28 Sep 2025 – 2 Oct 2026.Source:GitHub.
Less
More

Guardian AI

Real-Time Threat Assessment & Grooming Pattern Detection

Real-time Async ML

Modular, asynchronous threat detection platform coordinating transformer classification, speech-to-text, and LLM analysis with real-time alerting.

Architecture Decisions:
Asynchronous service-oriented backend on FastAPI, PostgreSQL, and WebSockets
Coordinated ML inference pipelines with background task workers for low-latency alerts
Speech-to-text integration with transformer-based multi-modal pattern detection
FastAPIPostgreSQLWebSocketsPyTorchReactML Pipelines
Source

Brain Tumor Classification System

Deep-Learning CE-MRI Classifier with Grad-CAM Localization

99.21% Accuracy

End-to-end medical computer vision pipeline with OpenCV preprocessing (denoising, CLAHE) and explainable diagnostic overlays.

Architecture Decisions:
Peak test accuracy of 99.21% across 1,519 CE-MRI evaluation images
DenseNet121 & ResNet50 architectures maintaining low latency of ~51ms per slice
Grad-CAM visual localization delivering explainable heatmaps via a Flask interface
PythonTensorFlow/KerasOpenCVFlaskDockerGrad-CAM
Source

QuizGenAI

AI-Powered Dynamic Assessment Platform — 2nd Prize Winner

🥈 2nd Place Hackathon

Modular backend services for dynamic quiz generation, answer evaluation, and real-time leaderboards with strict API contracts.

Architecture Decisions:
Structured FastAPI service architecture with PostgreSQL, SQLAlchemy, and token authentication
Isolated LLM services behind an abstraction layer for predictable structured outputs
Awarded 2nd Prize at the National Level Hire-4-Thon Hackathon (2026)
FastAPIPostgreSQLSQLAlchemyReactLLM LayerJWT Auth
Source

QuantNiti

Regime-Adaptive Quantitative Intelligence & Decision Support

Regime-Adaptive

Automated pipeline for regime-adaptive financial data modeling, time-series indicators, and systematic decision support.

Architecture Decisions:
Automated data ingestion and preprocessing workflows for multi-asset market regimes
Signal generation and decision-support algorithms with statistical validation
Modular Python design utilizing uv for reproducible dependency environments
PythonuvQuantitativeTime-SeriesSystems Design
Source
03 // SIGNALS & CREDENTIALS

Proven Engineering Signals & Foundations

Competitive global rankings, national hackathon podium, and high academic performance.

TryHackMe Global Ranking

TryHackMe Top 5% Globally

Ranked in top 5% worldwide solving CTFs with emphasis on web application vulnerabilities, authentication mechanics, and exploit analysis.

Field: Application Security
National Hackathon Runner-Up

Hire-4-Thon Hackathon — 2nd Place

Secured 2nd place nationwide in 2026 for building QuizGenAI, dynamic assessment generation backend powered by asynchronous FastAPI pipelines.

Year: 2026
Academic Record

KS Institute of Technology

B.E. in Computer Science & Engineering. Maintained an 8.88 CGPA through Semester 6 with strong command over algorithms, operating systems, and networking.

CGPA: 8.88 / 10.0Expected 2027