Open to GenAI roles · Hyderabad, India

Adarsh Pandey
builds AI that acts.

GenAI Engineer /

I design agentic systems that plan, retrieve, and execute — multi-agent workflows with LangGraph, session-isolated RAG pipelines, and production FastAPI services with tracing and observability built in.

Get in touch ↗Resume
01

About

design sensitivity × engineering depth

I work at the intersection of LLMs and systems engineering — turning models into agents that plan, call tools, retrieve knowledge, and ship real outcomes.

Right now that means building agentic AI at Techolution: rule ingestion via instant learning, tool-calling orchestration, and FastAPI services with full logging and tracing — so agents adapt to changing business rules without retraining.

I care about the unglamorous parts that make GenAI production-grade: session isolation, observability, evals, and clean APIs.

// bio

educationB.Tech CSE, IIIT Ranchi · 8.08 CGPA · 2026
currentlyAssociate AI Python Engineer, Techolution
focusAgentic AI · RAG · LLM orchestration
highlightAIR 2 / 25k · SourceCode 2024, IIT KGP
communityCoordinator, CP Wing · House of Geeks
locationHyderabad, India · Open to relocation
02

Selected Work

agents in production, not demos
P.01 — JUL 2025

Idea2Venture

Agentic AI system for end-to-end startup analysis

  • Turns a one-line startup idea into a full business plan — market validation, competitor analysis, pricing, MVP design, and risk assessment — autonomously.
  • Stateful multi-agent workflow on LangGraph: specialized agents run sequentially, share structured state, and selectively invoke web search and RAG tools.
  • Global-yet-isolated RAG on FAISS + MiniLM embeddings — every retrieved chunk is tagged with an analysis ID so concurrent requests never cross-contaminate.
idea→planner→researcher→analyst→critic→plan.json
FastAPILangGraphOllamaFAISSPostgreSQLRAG
View code ↗
P.02 — JUN 2025

ZeroCloud AI

Privacy-first, fully local AI assistant with RAG

  • A local-first AI assistant that runs on consumer hardware — sensitive-data workflows with PDF ingestion and retrieval, zero cloud LLM APIs.
  • Modular FastAPI backend with secure auth, API-key validation, and MongoDB persistence for session chat history and usage tracking across restarts.
  • Session-isolated RAG pipeline: PDF upload → extraction → configurable chunking/overlap → semantic search with MiniLM + FAISS, strictly scoped per session.
pdf→chunker→embeddings→FAISS→LLM (local)→answer
FastAPIOllamaNext.jsMongoDBFAISSMiniLM
View code ↗
03

Experience

shipping since day one
Jul 2026 — Present
● Current

Techolution Associate AI Python Engineer · Agentic AI, GenAI, FastAPI

  • Built rule ingestion via instant learning so agents adapt to evolving business rules in real time — no retraining, no code changes.
  • Engineered an agentic LLM system orchestrating complex workflows through tool calling and API integrations.
  • Shipped high-performance FastAPI services for agent lifecycle management, with refined system prompts and comprehensive logging + tracing for observability.
Jan 2026 — Jun 2026

Techolution AI Python Intern · Agentic AI, GenAI, FastAPI

  • Built agentic AI systems that autonomously plan, reason, and execute complex workflows through tool calling and API integrations.
  • Improved LLM accuracy using prompt engineering, grounding, and structured outputs, reducing hallucinations by up to 40%.
  • Developed self-healing AI workflows capable of automatic validation, error detection, and recovery for reliable autonomous execution.
Jul 2025 — Dec 2025

Xurge AI Python Developer Intern · Python, Next.js

  • Automated data pipelines (Selenium + Python) extracting and consolidating unstructured web data from 5+ sources — cut manual processing time by 80%.
  • Built tracking pipelines with Brevo and PostHog to monitor user onboarding and behavior.
  • Delivered an analytics pipeline with feature flags and cohort-based experimentation for data-driven decisions.
04

Stack

tools that ship agents

AI / GenAI — the core

Agentic AI SystemsRAGLLM OrchestrationMulti-Agent WorkflowsLangGraphLangChainVector DatabasesFAISSOllamaEmbeddings (MiniLM)Prompt EngineeringTool Calling

Languages

PythonSQLC++JavaScript (ES6+)TypeScript

Frameworks & Libraries

FastAPIFlaskPydanticNext.jsReactNode.jsSelenium

Databases & DevOps

MongoDBPostgreSQLMySQLORMGitHub ActionsCI/CD

Platforms & Tools

GitGitHubPostmanClerkPostHogBrevo
05

Proof of Work

ranked, rated, repeated
AIR 2
Among 25,000 participants in SourceCode 2024
Kshitij · IIT Kharagpur
2nd
UI/UX Arena 2025 — design & coding contest
BIT Mesra
1850
Knight on LeetCode · 450+ problems solved
LeetCode
1400 / 1800
Specialist on Codeforces · 4★ on CodeChef
Competitive Programming
06

FAQ

asked often, answered once

Because I've built the hard parts already — multi-agent orchestration with shared state, RAG with strict session isolation, tool-calling agents in production at Techolution. Not notebooks and demos: FastAPI services with auth, tracing, and observability.

An agentic startup-analysis engine (LangGraph, sequential specialized agents, RAG with analysis-ID tagging to prevent cross-contamination) and a fully local privacy-first assistant (Ollama, FAISS, session-scoped PDF RAG). Plus agent lifecycle services and instant-learning rule ingestion at work.

Yes — Next.js, React, and Node on the front, FastAPI and MongoDB/PostgreSQL behind. I can own an AI feature end to end: the agent runtime, the API, and the interface users actually touch.

I have completed my B.Tech at IIIT Ranchi (2026) and currently working as an Associate AI Python Engineer at Techolution in Hyderabad. Open to GenAI-focused roles and collaborations. The fastest way to reach me is email.

07 — Contact

Let's build something
that thinks.

Got a GenAI problem, a role, or just want to talk agents? My inbox is open.

adarshp.1133@gmail.com
GitHubLinkedInX / TwitterInstagram