Case File No. 001 · Status: Open

Tarun Singh
Chauhan

AI Engineer — LLM Systems · RAG · Multi-Agent Pipelines

Gotham has its detective. Production AI needs one too. I build the systems that catch what others miss — evaluation harnesses, red-teams, autonomous agents — engineered to hold up long after the demo ends.

Batman illustration
Subject: Verified
Scroll the file
— The Only Rule —

Code that fails silently is code that fails everyone.

— Alfred, if he shipped to production
The Utility Belt

Field Equipment

Five compartments. Every one earns its place — no gadget carried that hasn't been tested under fire.

LLM Engineering
01 / Signal

LLM Engineering

Prompt systems and typed contracts that hold up under production pressure, not just demo day.

OpenAIAnthropicOpenRouterLangChain
RAG and Retrieval
02 / Grapple

RAG & Retrieval

Vector search built to survive drift — with detection that fires before quality quietly decays.

QdrantEmbeddingsHit@kMRR
Multi-Agent Systems
03 / Squad

Multi-Agent Systems

Agents that check each other's work, orchestrated with typed state so failures can't slip through silently.

LangGraphFastAPIDocker
Evaluation and Red-Team
04 / Detector

Evaluation & Red-Team

Catching regressions and adversarial breaks before Gotham — or a customer — finds them first.

ROUGE-LBERTScoreCohen's κ
Infrastructure
05 / Frame

Infrastructure

The plain, unglamorous scaffolding underneath every gadget above — the part that actually ships.

DockerRedisPostgreSQLCI/CD
Training Record

Under the Cape

Before the case files, there's the training. Every detective starts somewhere.

JUN — AUG 2025

AI Engineer Intern

ZIDIO Development · Remote
  • Built AI-powered Python automation pipelines integrating LLM APIs across 5+ data sources, cutting downstream errors by ~35–40%.
  • Shipped prompt-engineering workflows and Power BI dashboards surfacing real-time KPIs, saving ~4–5 hours per weekly cycle.
JUL — AUG 2024

Generative AI & ML Intern

Linux World Pvt. Ltd. · Jaipur
  • Engineered ML data pipelines and applied generative AI across two projects: an NLP chatbot and a regression-based price predictor.
Batman illustration
Personnel File — Attached
Field Reports

The Case Files

Five investigations, five case files pinned to the board. Pull the thread on any one.

CASE-01 · SOLVED
CodePulse case photo

CodePulse

Cinematic, line-by-line AI code walkthroughs — LLaMA 4 via Groq returns a structured script driving an animated timeline with voice narration, variable tracking, and a live call stack.

Next.js 14Groq · LLaMA 4Web Speech API
Break Mode: AI bug-injection + fix narration
CASE-02 · SOLVED
LLM Evaluation Harness case photo

LLM Evaluation Harness

Benchmarked GPT-4o and Claude Sonnet across 50 MMLU prompts with ROUGE-L, BERTScore and bootstrapped 95% CIs — cutting regression detection from days to under 4 minutes.

LangGraphMLflowLangSmith
20+ adversarial red-team attack patterns
CASE-03 · SOLVED
Multi-Agent Code Review case photo

Multi-Agent Code Review

A 3-agent LangGraph pipeline — static analysis, OWASP scanner, LLM fix proposal — autonomously flagged 23 issues including 4 critical vulnerabilities.

LangGraphOWASPFastAPI
Per-review cost budget enforced at $0.50
CASE-04 · SOLVED
RAG Ops Platform case photo

Real-Time RAG Ops Platform

Production RAG with Qdrant vector search and cosine-similarity drift detection that auto-triggers re-indexing, plus a live ops dashboard tracking p95/p99 latency.

QdrantChart.jsRedis
75% cache hit rate on repeated queries
CASE-05 · SOLVED
Financial Research Agent case photo

Autonomous Financial Research Agent

A tool-augmented agent running a 4-step reasoning chain — market context to investment thesis — with a full SHA-256 reproducibility audit trail.

Yahoo Finance APIRedisDocker
Full reports generated in under 60 seconds
One Last Thing

Signal the Bat

Got a problem worth solving — a system that needs to actually hold up? Shoot your shot.

Send the Signal