Backend & AI Systems Developer

Savitender Singh

I build scalable backend systems and AI reliability pipelines focused on verification, security, and real-world utility.

AI RELIABILITY
HALLUCINATION DETECTION
BACKEND SYSTEMS
RAG PIPELINES
LLM VERIFICATION
SCALABLE APIs
SYSTEM SECURITY
INTEROPERABILITY
AI RELIABILITY
HALLUCINATION DETECTION
BACKEND SYSTEMS
RAG PIPELINES
LLM VERIFICATION
SCALABLE APIs
SYSTEM SECURITY
INTEROPERABILITY
// background

Verifiable Reliability First.

I'm a backend and AI systems engineer focused on building infrastructure that makes AI outputs trustworthy, traceable, and production-ready.

My work spans LLM reliability pipelines, healthcare interoperability systems, and scalable API backends, always with an emphasis on architecture first. I specialize in claims verification mechanisms to isolate and prevent LLM hallucinations, ensuring outputs comply with target guidelines.

CURRENTLY: Open to Full Stack / Backend / AI / PD roles · Building verification primitives

savitender@core: ~
>
// case studies
01 / FEATURED ARCHITECTURE

youFRHALLUCINATION MITIGATION

"Reliability layer for LLM outputs, claim extraction, verification, and explainability pipeline."

AI · NLP · spaCy · Flask · RAG · LLM Evaluation
THE PROBLEM

Large language models produce fluent but factually unreliable outputs. Hallucinations, which are confident false claims, make LLMs unsafe for high-stakes domains like healthcare, legal, and financial applications.

THE SOLUTION

Built a multi-stage reliability pipeline that intercepts LLM outputs and extracts atomic factual claims using custom NLP, verifies each claim against trusted knowledge bases, scores confidence weights, and highlights flagged claims.

// CHALLENGES & RESOLUTIONS
  • Decomposing complex compound sentences into atomic verifiable claims without losing semantic intent.
  • Optimizing RAG context lookups to achieve sub-second operational bounds under strict precision checks.
OUTCOME // QUANTIFIABLE SUCCESS

Successfully developed a drop-in interceptor pipeline enabling live fact-checking, auditing, and explainability for LLM applications.

// System Pipeline Architecture
01. LLM Raw OutputUnstructured text output02. Claim ExtractorspaCy NLP TokenizationExtracts atomic facts03. RAG VerificationRetrieval checksVector similarity lookup04. Scorer EngineProbability metricsWeights fact precision05. Explainer HubHallucination flagsHighlights core anomalies06. Audited ResponseTraceable outputsReady for production utility
[INPUT: LLM TEXT]// STAGE VERIFICATION INTERCEPTOR CLIENT RUNNING[OUTPUT: TRACEABLE FACT SHEET]
02 / INTEROPERABILITY NODE

MedFlow

"End-to-end patient referral and resource tracking system for distributed hospital networks."

Healthcare · Node.js · MongoDB · QR · REST APIs
THE PROBLEM

Independent regional hospitals lack unified stateful patient transfer systems, causing extreme delays, clinical communication failures, and resource bottlenecks during transitions.

THE SOLUTION

Developed an end-to-end patient referral workflow engine connecting hospital nodes, incorporating QR-based lookup grids for bedside patient updates, and real-time dashboard resource syncs.

// CHALLENGES & RESOLUTIONS
  • Designing schemas supporting isolation rules for multi-tenant clinical logs without cross-contamination.
  • Modeling clinical state machines (pending > accepted > closed) to handle transfer conflicts gracefully.
OUTCOME // QUANTIFIABLE SUCCESS

Functional healthcare backend implementing rigorous domain sync constraints and high-speed bedside QR routing.

// Node Interoperability Flow
Hospital Node ARest API InteroperableHospital Node BReferral DispatcherHospital Node CEmergency PortalCentral APICore RoutingNode.js / Express GatewayMongoDB InstanceMulti-tenant SchemaMulti-node tracking dataQR Bed bedside lookupSub-second QR lookupReferrals State MachinePending > accepted > closed
[MULTI-TENANT HOS NODES]// REST API CENTRAL COORDINATOR ROUTER SYSTEM RUNNING[QR DISPATCH ENGINE ACTIVE]
03 / APPLICATION BACKEND

Internship & Placement Portal

"Full-cycle recruitment backend for institutional placement workflows."

Backend · Auth · Node.js · MongoDB · REST APIs
THE PROBLEM

Placement cycles rely on manual coordinating and spreadsheets, introducing data corruption, secure validation failures, and massive delays for job candidates.

THE SOLUTION

Constructed a multi-role recruitment workflow engine incorporating signature-checked JWT authorizations, input sanity validations, and highly optimized database collections.

// CHALLENGES & RESOLUTIONS
  • Designing multi-tier JWT validations to guarantee robust access isolation between Students, Companies, and Administrators.
  • Achieving high write stability during flash registration hours through compound database indices.
OUTCOME // QUANTIFIABLE SUCCESS

Sturdy enterprise-level recruitment router backend featuring production-style validation structures.

// Authorization Routing Workflow
Role: StudentSubmit resumes & applicationsRole: CompanyPost listings, filter applicantsRole: AdminAudit placement workflowsJWT Auth FilterNode.js MiddlewareValidates claims & signatureMongoDB CollectionsRecruitment pipelinesStudent/Job schemas
[ROLE SPECIFIC CLIENT REQUESTS]// REST AUTHORIZER GATEWAY FILTERS ACTIVE[PIPELINE PERSISTENCE STACKED]
// core capabilities
// AI & Machine Learning
RAGLLM EvaluationHallucination DetectionNLPspaCyPrompt Engineering
// Backend Engineering
FlaskFastAPIDjangoNode.jsExpressREST APIsJWT AuthWebSockets
// Frontend
ReactReact NativeNext.jsTailwind CSS
// Databases
MongoDBPostgreSQLRedis
// DevOps & Tools
GitDockerVercelLinuxPostman
// track record
MAY 2025 - JULY 2025

Python Developer Intern@Shree Guru Gobind Singh Tricentenary University (SGTU)

  • Built backend for an Internship & Placement Portal with focus on scalability and security.
  • Developed RESTful APIs for authentication, profiles, resume uploads, and job applications.
  • Integrated MongoDB for flexible student data storage and seamless CRUD operations.
  • Implemented session-based authentication and validation for secure, reliable access.
// Skills:Python FlaskPython DevelopmentPython Developer InternRest API DesignFast APIFlask

Let's build something.

// Available for Full Stack, Backend, AI, and PD positions.