Ashish Verma
Let's chat →
Available for AI engineering & agent builds

Building autonomous AI agents that ship real work.

I'm Ashish — an AI engineer, senior Drupal developer, and full-stack architect. I design and deploy production agent systems, multi-model orchestration harnesses, and LLM-native SaaS — on top of 20+ years of enterprise Drupal and full-stack delivery for Fortune 500 clients.

See AI work → Background
Fremont, CA · SF Bay
20+ yrs engineering
Agents · LLMs · GCP Vertex / ADK
Autonomous agents Multi-model orchestration Vertex AI · ADK Production LLM systems Plan Mode Evals & guardrails RAG & embeddings On-device coding agents Autonomous agents Multi-model orchestration Vertex AI · ADK Production LLM systems Plan Mode Evals & guardrails RAG & embeddings On-device coding agents
7+
Production AI agents shipped
20+
Years engineering software
F500
Enterprise clients delivered
1
Live AI SaaS · revenue-generating
AI Engineering & Autonomous Agents

AI work, shipped to real users.

Production agents and LLM systems — built end-to-end, from prompt design and tool calling to deployment, evaluation, and safety hardening.

01
Live · Revenue-generating CLM SaaS Multi-tenant Word add-in

AI-Powered Contract Lifecycle Management Platform

An AI-native CLM SaaS where every step — drafting contracts inside Microsoft Word, clause libraries, obligation tracking, risk assessment — is powered by AI and tailored per tenant. In production with paying customers.

  • Automatic extraction of clauses, obligations, and key terms — captured into the tenant's library for reuse
  • Microsoft Word add-in that drafts whole contracts inside Word using each tenant's data
  • AI-suggested redlines, risk surfacing, and fallback clause recommendations
  • Multi-tenant SaaS with strict per-tenant data isolation — every tenant's CLM compounds in value
LLM contract understanding · clause embeddings · multi-tenant SaaS · Office.js add-in · per-tenant knowledge stores
02
Google Cloud Vertex AI Agent Engine Gemini + ADK

Enterprise AI Engineering on Google Cloud

Primary AI Engineer on a client engagement building production multi-channel customer interaction agents on Google Cloud's full enterprise agent stack — design through evaluation, guardrails, and hallucination detection.

  • Designed agent logic with Google's Agent Development Kit (ADK)
  • Grounded responses via function/tool calling + RAG
  • Vertex AI Agent Engine deployment; MCP servers and A2A for multi-agent coordination
  • ADK Evals, guardrails, hallucination detection, output validation
Gemini · Vertex AI · ADK · A2A · ADK Evals · MCP · Dialogflow CX · Cloud Run · Python · FastAPI
03
Coding Agent Jira × GitHub

Autonomous Coding Agent — Jira × GitHub

An AI software engineer that picks up Jira tickets the moment they're created, writes the fix on a new branch, raises a clean PR, and updates the ticket back with progress.

  • End-to-end fix workflow — branch, code, PR, status back to Jira
  • Plan Mode — multi-agent debate between ChatGPT, Claude, and other models for complex tickets; humans steer or veto mid-conversation
  • Codebase-aware context retrieval before any code is written
  • Routine bugs handled with no engineer involvement — cycle shrinks from days to hours
Multi-agent orchestration · Jira/GitHub APIs · codebase RAG · multi-model debate
04
Coding Agent GitLab Self-Hosted

Autonomous Coding Agent — GitLab Edition

GitLab-native version of the same agent — for teams running their full DevOps lifecycle inside GitLab, including self-hosted and on-prem deployments where no code can leave the environment.

  • Native integration with GitLab issues, branches, and Merge Requests
  • Works on gitlab.com and self-hosted — strict data and IP boundaries
  • Same Plan Mode multi-agent reasoning for complex tickets
  • Auto-generated MR descriptions and inline comments tied back to the issue
GitLab API · self-hosted deploy · codebase RAG · multi-model planning engine
05
Local Coding Agent Air-Gapped Ready

Local Development AI Agent — On-Device

Runs entirely on the developer's machine — watching for changes and failures, generating tickets, writing fixes, and running the full test suite before anything leaves the laptop.

  • Runs fully on-device — no code leaves the machine; works offline / air-gapped
  • Watches local changes, build and test failures in real time
  • Auto-generates Jira tickets, writes fixes locally, runs tests before declaring done
  • Deep IDE integration (VS Code, JetBrains); project memory learns team conventions
On-device LLM · IDE plugin · sandboxed execution · local file watcher · Jira integration
06
Marketing AI Creative Generation Social Publishing

Autonomous Marketing Agent — Ad Creation & Publishing

An AI agent that takes a marketing brief in plain English, generates a high-quality on-brand ad using the company's own assets, publishes across social channels after approval — and learns from past performance to recommend what should run next.

  • Brief-to-ad in plain English; pulls from brand database (product images, voice, campaigns, audience data)
  • Multi-platform publish — Instagram, Facebook, LinkedIn, X — with human approval before anything goes live
  • Performance-aware learning loop — every campaign sharpens the strategy for the next
  • Cuts routine campaign turnaround from days to minutes
LLM creative generation · multi-modal image gen · Meta/LinkedIn/X APIs · performance feedback loop
07
● In active development Custom LLM India-specialized

Nyaya — Indian Legal Domain LLM

A domain-specialized LLM for India's legal landscape — Constitution, Central and State Acts, the new criminal codes (BNS, BNSS, BSA), and procedural law — built for production deployment with government and enterprise adoption.

  • Open-weight base (Llama 3.x family), continued pre-training on the Indian legal corpus — strongest quality-to-cost ratio at 7B–13B parameters
  • Long context window (128K+) — non-negotiable for legal documents
  • Multilingual: English, Hindi, and major regional languages
  • On-prem / VPC deployment, India data residency, DPDP / ISO 27001 / CERT-In compliance pathway
Llama 3.x · continued pre-training · SFT + DPO · H100/A100 cluster · vLLM serving · Ray / Kubernetes
Capabilities

Stack across the AI and product layer.

Hands-on across the whole pipeline — agent design, LLM orchestration, RAG, evals — and the full-stack foundation that ships and scales it.

AI / LLM Engineering

Agent designTool callingRAG Multi-agent orchestrationPlan / debate protocols Evals & guardrailsHallucination detection Prompt engineeringEmbeddings Fine-tuning (SFT / DPO)Continued pre-training

Models & Platforms

Gemini EnterpriseVertex AI Agent EngineGoogle ADK Agent2Agent (A2A)MCP servers Dialogflow CXOpenAIClaude Llama 3.xvLLM / TGIADK Evals

Backend & Languages

Python (async · FastAPI)Node.js PHP · Drupal 7/8/9/10Laravel WordPressDjango REST / GraphQL

Frontend & Mobile

ReactAngularReact Native FlutterHTML5 / CSS3 / JS SASS / TailwindOffice.js (Word add-in)

Cloud & DevOps

Google Cloud (Vertex · Cloud Run)AWS AzureDigitalOcean / Linode Docker / KubernetesRay Cloud Trace / Monitoring / LoggingCI/CD

Data, Marketing & Integration

Versioned data lakesPer-tenant knowledge stores Salesforce Marketing CloudSalesforce CRM MarketoMeta / LinkedIn / X APIs
Career

20+ years building enterprise software.

Long arc from enterprise full-stack delivery into AI engineering — Fortune 500 clients, agile squads, and production systems people depend on.

AI Engineer · Senior Full-Stack Architect

Ai Tech

2005 — Present · 20+ years

  • Building production AI agents and LLM systems for enterprise and SaaS clients
  • Embedded as primary AI Engineer on Google Cloud enterprise agent engagements
  • Led enterprise-level Drupal and full-stack projects for Fortune 500 corporations
  • Architected complex e-commerce solutions with custom payment gateways
  • Owned full project lifecycles — requirements through deployment, evals, and ongoing ops

PHP / Drupal Developer

Enabling Dimensions, Delhi

Sep 2010 — Sep 2011

  • Developed custom Drupal modules and themes for client projects
  • Collaborated with design teams on responsive web solutions
  • Optimized website performance and implemented SEO best practices
Selected Web Projects

Enterprise builds, before the AI chapter.

A sample of long-running enterprise full-stack work — the foundation all of the AI work sits on top of.

IEEE Main Website

  • Custom modules for third-party service integrations
  • Advanced theming and frontend optimizations
  • Scalable architecture for a high-traffic enterprise site
Visit site →

Emory School of Public Health

  • Acquia Site Studio rollout for content efficiency
  • Custom data migration scripts
  • Responsive frontend with advanced functionality
Visit site →

Quantum Expeditions

  • Drupal 8 → 9 migration
  • React components for dynamic trip booking
  • Multi-currency support; REST APIs for real-time updates
Visit site →

Israel Museum of Jerusalem

  • Responsive museum site with Bootstrap, Angular, CSS3
  • Custom filtering and multi-site templates
Visit site →

University of Southern California

  • WordPress multisite architecture
  • Custom educational themes
  • Content workflows for multiple departments
Visit site →

IAMS Pet Food

  • Acquia Site Studio for streamlined CMS
  • Custom API integration for marketing & analytics
  • End-to-end Drupal 9 migration with zero downtime
Visit site →

Have an AI agent or LLM build in mind?

I take on AI engineering, agent design, and production LLM systems — alongside the full-stack work to make them ship and scale.

Let's chat → WhatsApp
FAQ

Frequently asked questions.

Quick answers about what I build, how I work, and where I'm based.

What does Ashish Verma do?

Ashish Verma is an AI engineer and senior full-stack architect based in the San Francisco Bay Area. He designs and ships production-grade autonomous AI agents, multi-model orchestration harnesses, and LLM-native SaaS products — and brings 20+ years of enterprise software delivery underneath the AI layer to make sure those systems actually ship and scale.

What AI services do you offer?

Autonomous AI agent design and engineering · multi-agent orchestration and Plan Mode style debate protocols · Retrieval-Augmented Generation (RAG) systems · production LLM deployment on Google Vertex AI, the Agent Development Kit (ADK), and Agent2Agent (A2A) · MCP server integration · evaluations, guardrails, and hallucination detection · custom LLM development including continued pre-training and fine-tuning (SFT / DPO) · AI-native SaaS architecture.

Which AI platforms and models do you specialize in?

Google Cloud's full enterprise stack (Gemini Enterprise, Vertex AI Agent Engine, Google ADK, Agent2Agent / A2A, ADK Evals, Dialogflow CX, Cloud Run), Anthropic Claude, OpenAI GPT models, and open-weight Llama 3.x family models served via vLLM and TGI for cost-effective on-prem deployment.

How many years of engineering experience do you have?

20+ years building production software, starting in 2005. The arc covers enterprise PHP, Drupal, and full-stack work for Fortune 500 clients (IEEE, Emory, USC, IAMS, Israel Museum of Jerusalem, and others) — and over the last several years, a sharp pivot into AI engineering, agent systems, and LLM platforms.

Where are you based and do you work with clients remotely?

Based in Fremont, California — San Francisco Bay Area — and work with clients globally. Comfortable embedding directly inside client engineering squads as an AI engineer, or operating as an independent build partner for greenfield AI products.

Do you still take on Drupal development work?

Yes. Drupal is the long-running foundation of the practice — 20+ years across Drupal 7, 8, 9, and 10 for Fortune 500 clients including IEEE, Emory School of Public Health, USC, IAMS Pet Food, Quantum Expeditions, and the Israel Museum of Jerusalem. Comfortable with custom module and theme development, large-scale Drupal version migrations (8 → 9, 9 → 10), Acquia Site Studio rollouts, multisite WordPress architectures, and custom API integrations for marketing and analytics platforms. Increasingly, Drupal engagements combine with AI work — Drupal sites augmented with AI agents, RAG over CMS content, and LLM-powered editorial workflows.

Can you build AI agents that run on-premise or air-gapped?

Yes. I've built local-first coding agents that run entirely on the developer's machine for IP-sensitive and regulated environments (finance, healthcare, defense). For server-side, I've shipped self-hosted GitLab-native agents and on-prem LLM deployments for clients with strict data residency requirements. Compliance pathways including DPDP, ISO 27001, and CERT-In are part of the toolkit.

What's the best way to get in touch?

Email ashish@ashtechy.com or WhatsApp +1 (341) 221-9451. I typically reply within one business day.

Contact

Get in touch.

✉
Email

ashish@ashtechy.com

ashish.arush85@gmail.com

☎
Phone / WhatsApp

+1 (341) 221-9451

Open WhatsApp →

⌖
Based In

3888 Invent Terrace
Fremont, CA 94539
San Francisco Bay Area