Hitesh Wadhwa

// about

I build systems that actually hold up in production. The kind that scale, fail gracefully, and don't wake someone up at 3am. Reliable and boring in the best way.

Most of my work has been in distributed systems, orchestration, and automation. I like taking messy, ambiguous problems and turning them into something clean: fewer edge cases, good observability, execution paths that just work.

I'm pretty stack-agnostic. I care more about understanding how a system behaves end-to-end than any specific tool or framework. Always tinkering on a side project too.

// quick facts

locationBay Area, CA

educationM.S. CS · UMass Amherst (3.97)

focusdistributed systems · AI infra

buildingWebChat browser extension

// experience

Stealth Startupa16z-backed● live

2025 – Present · Bay Area, CA

Founding Member of Technical Staff

Building computer use agents for enterprise. Autonomous browser and desktop automation at scale. Founding technical team member shaping architecture, product, and culture from day zero.

Computer Use AgentsEnterprise AIFounding Engineer

Articul8 AI

Aug 2024 – Apr 2025 · Dublin, CA

Senior Software Engineer → Software Engineer

Architected a Ray-based agentic ingestion platform with IO-parallel execution, cutting p95 latency 9× from minutes to seconds at multi-tenant scale.
Led a 7-engineer cross-functional program delivering a cloud-agnostic marketplace platform across entitlements, metering, and billing.
Built Knowledge Graph APIs enabling intelligent entity linking, powering Essential Platform launch across 10+ internal teams.
Owned end-to-end auth architecture using API gateways and token-based enforcement for external agents.

RayGraphQLKubernetesKnowledge GraphRabbitMQ

Microsoft Research

Feb 2024 – May 2024 · Cambridge, MA

Graduate Researcher — LLMs

Co-authored an arXiv paper on contextual shortcuts in RAG. Studied how LLMs prioritize retrieved context over parametric memory across Phi-2 (2.7B) and LLaMA-2 (7B).
Ran 50+ causal tracing experiments; observed up to 5× decrease in parametric reliance in RAG vs. vanilla settings.

RAGLLMsPyTorchCausal Tracing

App Orchid Inc.

Jul 2023 – Sep 2023 · San Ramon, CA

Data Science Intern

Applied NLP (BERT, FastText, SpaCy) to automate column extraction for a knowledge-graph pipeline, saving 20+ hours/week.
Increased labeling accuracy by 25% via fine-tuned SpaCy NER models.

NLPBERTSpaCy

// projects

WebChatIn Progress

A browser extension that lets you perform actions on any website using natural language chat. Describe what you want to do — WebChat handles the rest.

Browser ExtensionTypeScriptLLM AgentsWeb Automation

github.com/hitesh-wadhwa ↗

YoutubeChatSide Project

ReAct-based RAG pipeline that lets you have a conversation with any YouTube video. Memory buffers maintain context across the session.

RAGLangChainPythonStreamlit

Slouch Detection SystemHardware + ML

End-to-end posture detection: ESP8266 and Raspberry Pi collect sensor data, a Flask backend runs a neural network for classification, and a React dashboard shows your sitting history.

IoTPythonFlaskReactNeural Network

Fault-Tolerant Trading PlatformDistributed Systems

Microservice architecture with caching, replication, and fault tolerance. Pyro5 for inter-service RPC, containerized with Docker, deployed across AWS EC2 with NGINX.

MicroservicesPythonDockerAWSFault Tolerance

// early builds · automation instinct

COVID-19 News Bot2020

Automated WhatsApp bot that scraped live COVID updates during lockdown. 500 people joined the group — built in a weekend, ran for months.

GRE WordCollector2021

Vocab tool with a scraper backend and custom learning algorithm. Spread across my college — peers used it daily while prepping for the GRE.

Hitesh
Wadhwa_

Writing is coming.

Let's build something
interesting.

HiteshWadhwa_

Writing is coming.

Let's build somethinginteresting.

Hitesh
Wadhwa_

Let's build something
interesting.