Pratheep S

AI Automation & Generative AI Engineer

I'm Pratheep.

I build AI systems that replace work your team shouldn't be doing manually. Two years of real client work. Actual pipelines in production — not tutorials. Open to full-time roles and project-based work.

2+ Years of real client work
3+ Industries served
0 Projects abandoned mid-way
100% Production deployments
About

What I do

I build AI systems that cut manual work, reduce operational overhead, and plug directly into tools your team already uses.

Over the past two years, I've worked directly with clients across legal, SaaS, and e-commerce — mapping their manual workflows, identifying where AI creates the most leverage, and shipping systems that run in production.

My work sits at the intersection of large language models, workflow automation, and backend integration. I don't just connect an API and call it done — I build the full pipeline, document it properly, and make sure whoever inherits it can actually maintain it.

I'm currently in my 6th semester at Bannari Amman Institute of Technology. The degree gives me the foundations. The two years of client work gave me everything else.

If you're a company looking for someone who can build, ship, and own AI systems — I'm available for full-time roles, internships, and project-based work.

What I bring to a team

  • Production AI systems, not prototypes
  • End-to-end ownership from design to deploy
  • Clear documentation and maintainable code

What I'm looking for

  • A team building real AI products
  • Problems worth solving at scale
  • Ownership and direct responsibility
Skills

Technical Stack

Generative AI & LLMs

LangChain LlamaIndex OpenAI API Anthropic Claude API Hugging Face RAG Pipelines Prompt Engineering Fine-tuning (LoRA / QLoRA) Vector Databases Pinecone Weaviate ChromaDB

AI Automation & Orchestration

n8n Make (Integromat) Zapier Multi-Agent Systems Tool-Use Agents Agentic Workflows API Chaining

Backend & Integration

Python FastAPI Flask REST APIs Webhooks PostgreSQL SQLite Firebase

Dev Tools & Infrastructure

Git GitHub Actions Docker Postman Supabase Notion API Google Workspace APIs
Projects

Selected Work

01 / 03 Legal AI

Intelligent Document Review System

Client: Legal Services Firm

Problem

A legal consultancy was spending 6–8 hours per week manually reviewing client-submitted documents — checking for missing clauses, inconsistent terminology, and compliance gaps before attorney review. The process was slow, error-prone, and bottlenecked their onboarding pipeline.

Solution

Built a document intelligence system that ingests uploaded contracts and legal briefs, runs them through a structured LLM pipeline, and generates a clause-by-clause review report — flagging risks, missing sections, and inconsistencies against the firm's compliance checklist.

How It Works

Documents are uploaded via a lightweight interface. A preprocessing layer chunks and embeds the text into a vector store. A RAG pipeline queries against the firm's internal compliance ruleset. The LLM generates structured JSON, rendered as a formatted review report ready for attorney sign-off.

Python LangChain ChromaDB OpenAI API FastAPI React

System Workflow

  1. Document UploadPDF / DOCX
  2. PreprocessorChunk + Clean
  3. Vector StoreChromaDB Embed
  4. RAG PipelineCompliance Query
  5. LLM ReviewOpenAI GPT-4
  6. JSON RendererStructured Output
  7. Review ReportAttorney Ready

Impact

  • Review time cut from 6–8 hours to under 25 minutes per week
  • Clause error catch-rate increased by approximately 40%
  • Attorney time fully redirected to billable work
02 / 03 Sales Automation

Automated Lead Qualification & CRM Enrichment Pipeline

Client: B2B SaaS Company

Problem

A B2B SaaS company was receiving 300–400 inbound leads per month across email, web forms, and LinkedIn. Their sales team was manually reading each submission, scoring the lead, and updating their CRM — consuming 3–4 hours daily with inconsistent scoring across reps.

Solution

Built an end-to-end AI pipeline that captures leads from all sources, enriches them with company data, scores them against an Ideal Customer Profile, and auto-populates the CRM with a structured summary and priority tag — zero manual handling for 85% of inbound volume.

How It Works

Leads are captured via webhook integrations across channels. An n8n orchestration layer routes each lead through an enrichment step using web scraping and company data APIs, then passes the enriched profile to an LLM that scores fit against the ICP and writes a one-paragraph sales brief. Output is pushed directly to HubSpot via API. High-priority leads trigger a Slack alert.

n8n Python OpenAI API HubSpot API Slack Webhooks Company Enrichment APIs

System Workflow

  1. Lead CaptureEmail / Form / LinkedIn
  2. Webhook Routern8n Orchestration
  3. Data EnrichmentCompany APIs + Scrape
  4. ICP ScorerLLM Evaluation
  5. Sales Brief WriterLLM Summary
  6. HubSpot CRMAuto-populated
  7. Slack AlertHigh-priority leads only

Impact

  • 85% of leads processed with zero manual work
  • Sales team response time: 24 hours to under 2 hours
  • Eliminated inter-rep scoring inconsistency entirely
03 / 03 Enterprise AI

Internal Knowledge Base Agent

Client: Mid-Size E-Commerce Operations Team

Problem

A growing e-commerce company had 4 years of SOPs, process documents, supplier contracts, and policy files scattered across Google Drive, Notion, and email threads. New hires spent their first two weeks asking senior staff basic operational questions. Senior staff were losing 45–60 minutes daily answering the same questions repeatedly.

Solution

Built a private internal AI agent that indexes the company's entire documentation corpus and answers operational questions in plain language — citing the exact source document and page so answers can be verified by anyone on the team.

How It Works

Documents from Google Drive and Notion are ingested on a scheduled sync via their respective APIs. Content is chunked, embedded, and stored in a vector database. A FastAPI backend handles queries from a simple internal chat interface. The LLM retrieves relevant chunks via semantic search, generates a precise answer, and appends source citations. New documents are automatically re-indexed on upload.

Python FastAPI LlamaIndex Pinecone Google Drive API Notion API OpenAI API

System Workflow

  1. Doc SourcesGoogle Drive + Notion
  2. Scheduled SyncAuto re-index on upload
  3. Chunk + EmbedLlamaIndex
  4. Vector StorePinecone
  5. Staff QueryInternal Chat Interface
  6. Semantic SearchRAG Retrieval
  7. LLM Answer Generation+ Source Citation
  8. Verified AnswerDelivered with doc reference

Impact

  • New hire ramp-up time: 2 weeks to 4–5 days
  • Senior staff reclaimed approximately 45 minutes per day
  • 1,200+ queries handled in first 60 days with zero escalations
Experience

Work History

AI Automation Engineer — Freelance (Independent)

2 Years · Ongoing  |  Remote  |  Legal, SaaS, E-Commerce

Worked directly with business owners and operations leads to identify high-cost manual workflows and replace them with AI-powered systems. Scope of work ranged from single-pipeline builds to multi-system integrations connecting LLMs, CRMs, document platforms, and communication tools.

Every project was scoped, delivered, and supported without a team — which means the systems are built to be maintainable, not just functional on demo day.

Scope of Work

Requirements discovery with non-technical stakeholders, system architecture and pipeline design, full-stack build and deployment, handover documentation, and post-launch iteration based on real usage.

Delivery Standard

Every build is documented, version-controlled, and designed to be handed off to an internal team without dependency on the original developer.

Client Profile

Founders, heads of operations, and product leads who needed AI systems built fast — without hiring a full team or waiting months for an agency.

Working Style

Async-first. Direct communication. No unnecessary meetings. Results communicated through working software, not status updates.

Client engagements have involved NDA-covered work. Detailed case studies available on request.

Education

B.Tech — Artificial Intelligence & Data Science

Bannari Amman Institute of Technology, Erode, Tamil Nadu

2023 – 2027  ·  6th Semester  ·  Currently Pursuing

The degree covers the theoretical foundations — machine learning, neural networks, data engineering, NLP. The parallel client work over two years covers the rest: real constraints, real data, real accountability.

Contact

Got something you want built?

If you're a company looking for someone who can own AI systems end-to-end — or if you have a workflow problem that's costing your team time and accuracy — let's talk.

I'm open to full-time roles, internships, and freelance projects. I don't do vague calls. You tell me the problem, I'll tell you exactly what I'd build.

Open to opportunities — full-time, internship, or project-based.