AI Software Engineer

AI Engineer building practical AI systems.

I design and build RAG systems, multimodal AI applications, and AI automation workflows that transform documents, images, emails, and business data into real business outcomes.

View projects

Discuss a project

GH IN X

RAG

Multimodal

Automation

Data AI

workflow pipeline

live

Documents · Emails · Images · CSV

source inputs

INPUT

Extract · Chunk · Analyze

pre-processing

PROCESS

LLM / Vision

model layer

Insights · Automation

business output

OUTPUT

Focus Areas

RAG • Vision • Automation

AI Workflows

RAG · Vision · Data

Tech Domains

CRM · Education · Ops

01 · what i build

AI Systems I Build

module

AI Knowledge Systems

Build grounded AI applications using retrieval, embeddings, and structured reasoning.

Ingest→Embed→Retrieve→Answer

LangChainOpenAIChromaDBSupabase

module

AI Data Platforms

Transform structured datasets into insights using natural language and executable analysis.

Parse→Reason→Verify→Report

FastAPIpandasOpenAI

module

Multimodal AI

Process documents and images using vision models and automated evaluation pipelines.

Capture→Encode→Fuse→Decide

OpenAIGeminiPython

module

AI Automation

Build AI-assisted workflows for CRM enrichment, email processing, and business operations.

Plan→Act→Observe→Loop

FirecrawlGeminiFastAPI

02 · featured projects

Selected engineering work

04 systems

01 / RAG · Document Intelligence2026

AskMyDoc

From PDF chat demo to reliable multi-user RAG product.

Architecture

Multi-user RAG Workspace

Output

Structured Grounded Answers

01 / RAG · Document Intelligence2026

AskMyDoc

From PDF chat demo to reliable multi-user RAG product.

Architecture

Multi-user RAG Workspace

Flow

Auth → Upload → Index → Retrieve → Validate

Output

Structured Grounded Answers

Started as a single-document RAG experiment and evolved into a multi-user document workspace with persistent conversations, ownership-aware access, structured answers, and evidence-aware validation.

Auth + Persistent Workspace
Structured Answers + Citations
Lifecycle-Safe Document Handling

Challenge

Turning a working RAG demo into a trustworthy product with strong contracts, ownership rules, and graceful failure handling.

FastAPILangChainChromaDBOpenAISupabaseNext.js

Case Study→Repo

02 / AI Data Platform2026

AskMyData

Ask CSV data questions and get executable analysis with visual output.

Architecture

CSV Upload

Output

Analysis + Charts

02 / AI Data Platform2026

AskMyData

Ask CSV data questions and get executable analysis with visual output.

Architecture

CSV Upload

Flow

Question → Generate Python → Execute → Visualize

Output

Analysis + Charts

Upload CSV files, ask questions in natural language, generate Python analysis code, execute it in a sandboxed flow, and return answers with charts and insights.

Sandboxed Execution
Session Recovery
Charts & Insights

Challenge

Executing LLM-generated analysis safely while preserving flexibility.

FastAPIOpenAIReactSupabasePostgreSQL

Case Study→Repo

03 / AI Automation · CRM2026

AI-Powered CRM Lead Enrichment

Automated CRM enrichment from inbound inquiry emails.

Architecture

Email Processing Pipeline

Output

CRM Intelligence

03 / AI Automation · CRM2026

AI-Powered CRM Lead Enrichment

Automated CRM enrichment from inbound inquiry emails.

Architecture

Email Processing Pipeline

Core Stack

Firecrawl + Gemini

Business Use

CRM Intelligence

Built an AI-powered CRM workflow that processed incoming inquiry emails, filtered fake and spam leads, extracted company websites, scraped business information using Firecrawl, and generated structured CRM intelligence using Gemini.

Email → Filter → Website → Firecrawl → Gemini → CRM
Lead qualification and spam filtering
Structured CRM enrichment output

Challenge

Filtering low-quality inquiries without rejecting legitimate business leads.

PythonFirecrawlGeminiOdoo

Client Project

04 / Multimodal AI · Education2026

Multimodal AI Exam Evaluation

Automated assignment grading with multimodal vision models.

Architecture

Vision Evaluation Pipeline

Output

Grade + Feedback

04 / Multimodal AI · Education2026

Multimodal AI Exam Evaluation

Automated assignment grading with multimodal vision models.

Architecture

Vision Evaluation Pipeline

Core Models

Gemini + OpenAI Vision

Output

Grade + Feedback

Built an automated grading workflow where student assignment images were evaluated against teacher answer keys and graded using multimodal vision models.

Assignment → Vision → Compare → Grade → Feedback
Image-based answer evaluation
Automated grading feedback loop

Challenge

Maintaining grading consistency across handwritten student submissions.

GeminiOpenAI VisionPython

Client Project

03 · stack

Engineering Toolkit

AI Engineering

OpenAI APILLMsLangChainRAG PipelinesChromaDBAI AutomationMultimodal AI

Backend

FastAPIDjangoPythonREST APIsPostgreSQLMongoDBQuartSupabase

Tools

GitGitHubDockerPostmanLinux

04 · career

Experience

Feb 2025 — Present

Software Developer · KNYSYS

Build AI-powered automation systems and backend platforms across CRM, education, and enterprise workflows. Worked directly with clients to gather requirements and translate business workflows into AI-enabled solutions.