Open to AI Backend / ML Systems roles

AI Backend Engineer,
LLM/RAG Systems Builder.

I build backend-heavy AI systems: async ingestion pipelines, RAG, vector search, model routing, observability, recovery workflows, and evaluation-driven ML infrastructure.

Get in touch View work Resume
7 AI systems
● Available · IST
Sonu
AI Backend Engineer
B.Tech CS · Manav Rachna · 2026
Java 21 Spring Boot Python FastAPI RAG LangGraph PostgreSQL Redis RabbitMQ Docker

Academic
Background

Computer Science graduate focused on backend-heavy AI systems: RAG, async pipelines, vector search, model routing, and production observability.

B.Tech Computer Science
Manav Rachna University · Faridabad
2022 — May 2026

Graduated May 2026. Focused on software engineering, data structures, backend systems, and self-directed ML engineering. Built projects around RAG, LLM infrastructure, distributed backend workflows, observability, and evaluation.

Backend Systems AI Systems RAG Distributed Systems

Featured AI Systems

A focused set of backend-heavy AI projects: agentic RAG, SQL agents, legal RAG infrastructure, documentation learning systems, LLM routing, and evaluated financial RAG.

GitHub
Legal RAG infrastructure

LexGuard

Production-oriented legal document RAG backend with async ingestion, Transactional Outbox, RabbitMQ, Cloudflare R2, pgvector HNSW search, API key auth, tenant-aware rate limiting, supervisor recovery, and Prometheus/Grafana observability.

● 8-service Docker stack · pgvector p99 0.48s concurrent
Java 21 Spring Boot Python FastAPI RabbitMQ pgvector Prometheus Docker
GitHub Demo
Enterprise agentic RAG

Enterprise Agentic RAG

Production-grade agentic RAG system with a LangGraph Planner → Retriever → Responder loop, NeMo Guardrails input filtering, Portkey fallback/retry/caching, Qdrant retrieval, FlashRank reranking, and Logfire + LangSmith tracing.

● 15 golden Q&A pairs · 6 RAGAS metrics · adversarial noise testing
LangGraph NeMo Guardrails Qdrant Portkey FlashRank RAGAS Logfire LangSmith
GitHub
Natural language SQL agent

SQL Database Agent

Natural-language-to-SQL agent that generates, validates, executes, and audits database queries. READ queries are allowed, WRITE operations require confirmation tokens, and DDL is blocked outright for safety.

● 29 tests passed · audit log · schema cache · DeepEval harness
FastAPI LangChain SQLAlchemy LiteLLM Groq Gemini Logfire DeepEval
GitHub Demo
Agentic learning system

VaultLearn

Agentic documentation learning system that crawls live docs, generates structured study plans, indexes pages into ChromaDB, runs module-scoped RAG sessions, detects learning struggles, and writes Obsidian-compatible memory notes for cross-session continuity.

● FastAPI docs: 152 pages · 6 modules · 3,493 chunks
LangGraph FastAPI ChromaDB CrossEncoder Playwright React Groq
GitHub Live
LLM infrastructure

LLM Cost Router

Domain-general LLM routing layer benchmarked on financial queries. Uses FAISS + Redis semantic cache, TF-IDF + Logistic Regression complexity routing, Gemini model tiers, timeout handling, fallback behavior, and benchmark telemetry.

● 93.19% synthetic savings · realistic production estimate 40–60%
Python FastAPI FAISS Redis scikit-learn Gemini Docker
GitHub Live Demo
Evaluated RAG pipeline

Financial RAG API

Financial RAG over Apple and Microsoft 10-K filings using IBM Docling ingestion, BM25 + dense hybrid retrieval, persistent ChromaDB, citation-aware prompts, Gemini 2.5 Flash generation, RAGAs evaluation, MLflow, and Prometheus.

● Context Recall 0→1.0 · Faithfulness 0.82 · 453+ chunks
Python FastAPI Docling ChromaDB BM25 Gemini RAGAs MLflow

Core
Skills

Clean, recruiter-friendly stack focused on widely used backend, ML, database, and deployment skills.

01
Programming Languages
Java Primary
Java 21
Java 21 Spring Boot REST APIs JPA / Hibernate OOP Concurrency
Python Primary
ML / Backend
FastAPI PyTorch scikit-learn Pandas / NumPy asyncio Pydantic
02
Backend Engineering
Spring Boot FastAPI REST APIs JWT/Auth WebSockets Microservices Background Jobs API Security
03
ML & AI Systems
Machine Learning Deep Learning RAG LLM Applications Agentic AI LangChain LangGraph Vector Search Embeddings Reranking Model Evaluation
04
Data & Messaging
PostgreSQL SQL SQLite Redis RabbitMQ SQLAlchemy Database Design
05
Infra & Observability
Docker Docker Compose Git GitHub Actions Linux AWS RDS Render Hugging Face Spaces Prometheus Grafana

Let's build
reliable AI systems.

Open to AI Backend Engineer, ML Systems, and backend-heavy LLM/RAG roles. If you are building useful AI infrastructure, I would love to talk.

Local time
—
—
Typical response
Within 24 hours
Open to
Full-time · Contract · Freelance
Work preference
Remote · All timezones
Currently building
Agentic RAG · SQL Agent · ML Engineering
Drop me a line