gitesh.poudel

AI Engineer · Data Scientist

Hi, I'm Gitesh Poudel.

I build LLM agents and data-driven models that are measured, not guessed: retrieval with citations, tools for exact numbers, and eval suites that say when something works.

Try the live AI demoSee projectsMy journey →

Selected work

Projects

2026-10live demo

Experimentation Copilot

An AI analyst for A/B tests that never does the math itself.

An MCP server that gives any AI client validated experiment statistics: power analysis, sample-ratio checks, CUPED, delta-method ratio metrics, sequential and Bayesian tests, and segment-level effects. An agent uses those tools to write ship / don't-ship readouts in which every number comes from a tool call.

DSA/B test design with pre-registered decision rulesCUPED, delta method, sequential and Bayesian tests
AIMCP server with 12 typed, read-only toolsTool-calling agent that cites every number to a tool result
Read the case study →Try it ↗Code ↗

98%

Decisions that match the pre-registered rule, 40 held-out simulated experiments × 3 runs

Agent with tools98%
Same model, summary stats only73%
2026-10live demo

EDGAR RAG Analyst

An AI agent for SEC filings that cites every claim and computes every number.

A RAG and function-calling agent over the 10-K filings of 15 public companies. Explanations come from hybrid search over filing text; figures come from SEC XBRL data through typed tools, and every number in an answer is checked against its sources. Every pull request must pass an eval gate in CI, and each answer is traced with its cost.

DSGolden-set evaluation with a held-out test splitPaired bootstrap confidence intervals over repeated runs
AIHybrid retrieval: BM25 + vector, fused by rankFunction calling over SEC XBRL data and a calculator
Read the case study →Try itCode ↗
97.5%vs 93.5% for vector search only (v1)
2026-10

Alpha Research Engine

Autonomous quant research for crypto, on local infrastructure.

Ingests market time-series and social sentiment, scores sentiment with local embeddings, computes rolling signal metrics, and backtests strategies with strict in-sample/out-of-sample separation. Optuna walk-forward tuning, LangGraph research agents, paper trading and a monitoring dashboard sit on top.

DSStrict in-sample / out-of-sample separationWalk-forward tuning with Optuna
AILangGraph research agentsSentiment from local embeddings (Ollama)
Read the case study →Code ↗

Open source

Latest on GitHub

All repositories ↗

Toolbox

Skills

AI Engineering

  • LLM agents & function calling
  • MCP servers
  • RAG & hybrid search
  • LLM evaluation
  • LLM observability (OpenTelemetry)
  • CI eval gates

Data Science & Experimentation

  • A/B testing & experimentation
  • Causal inference & uplift
  • Sequential & Bayesian testing
  • Variance reduction (CUPED)
  • Monte Carlo simulation
  • NLP & BERT

Data & Infra

  • Python, SQL
  • Snowflake & dbt
  • DuckDB, LanceDB, Qdrant
  • Pydantic, asyncio
  • GitHub Actions, pytest
  • Docker, AWS

Let's build something.

Open to collaborations and roles in AI engineering and data science.

GitHubLinkedInEmail