Akash Chauhan — Production RAG & AI-Agent Systems

I build production RAG & AI-agent systems for B2B SaaS.

Shipped and evaluated — not demos. RAGAS-scored retrieval, grounded agents, deployed and audit-grade. Fixed scope. Two weeks. Done.

92/22,000

Cognizant national rank

1.0

RAGAS context precision

AI systems shipped live

VS Code

Published extension

Real systems. Real evals. Shipped to production.

Not toy chatbots. Retrieval pipelines with measured precision, multi-agent orchestration, and separation of concerns that holds up under audit.

GreenLedger-AI

RAG · AWS Bedrock · FastAPI

Problem

ESG reporting needs legally audit-grade numbers from messy documents — but raw LLMs hallucinate math and can't be trusted with compliance figures.

Built

A RAG pipeline on AWS Bedrock Titan v2 embeddings (FAISS, cosine) with strict LLM/math separation — Claude extracts raw values only, Python owns every GHG, water and POSH calculation.

context_precision 1.0

context_recall 1.0

faithfulness 0.86

input tokens −40%

React · Node.js · Python FastAPI · AWS Bedrock · BullMQ · AWS S3 · FAISS

Placement Pilot

Multi-agent · RAG · Real-time

Problem

Placement cells react too late to at-risk students, and generic LLM advice isn't grounded in a student's actual resume or placement data.

Built

A custom ReAct multi-agent system (Recon, Strategy, Sentinel) — no LangChain. Recon queries a per-user FAISS store before reasoning; Sentinel autonomously monitors risk scores and fires real-time alerts.

per-user FAISS

RAGAS-logged faithfulness

live alerts Socket.io

Node.js · React · Redis · BullMQ · Socket.io · Ollama · MongoDB

Flashen

LLM pipeline · Spaced repetition

Problem

Turning dense PDFs into effective study material is manual and slow, with no retention tracking.

Built

PDF → flashcard decks via Groq LLM with SM-2 spaced repetition, mastery analytics and retention heatmaps. Hardened: JWT, Helmet, CORS allowlist, route-level rate limits. Validated with Playwright E2E + k6 load tests.

tested Playwright E2E

load k6

scheduling SM-2

React 19 · Node.js · MongoDB · Groq · Playwright · k6

One thing, done right.

A full-time AI engineer is ₹40L+/yr and months to hire. This is fixed-scope, fixed-price, and shipped in two weeks.

Productized · Fixed scope

Production RAG chatbot on your docs — evaluated and deployed.

Price

₹2.5L / $3K

Timeline

2 weeks

Includes

Eval + Deploy

RAG pipeline on your documents — chunked, embedded, retrieval-tuned
RAGAS evaluation report — precision, recall, faithfulness measured, not promised
Deployed and documented — handover so your team can own it
Grounded answers with citations, not hallucinated guesses

Start a project →

Why trust me with it.

I'm Akash — an engineer who builds production AI, not proofs of concept. I've shipped four live AI systems, published a VS Code extension on the official marketplace, and ranked 92 of 22,000 nationally at Cognizant.

At Uplyx Solution I built fault-tolerant REST APIs that cut response latency 15%, set up CI/CD with Jenkins and Docker, and ran scalable hosting on AWS (S3, CloudFront, Route53). I care about evals, failure modes and observability — the boring things that make AI actually trustworthy in production.

I build production RAG & AI-agent systems for B2B SaaS.

Real systems. Real evals. Shipped to production.

One thing, done right.

Production RAG chatbot on your docs — evaluated and deployed.

AI in the spotlight. Full-stack as the muscle.

Why trust me with it.

Got docs that need an AI brain? Let's ship it.