Need AI That Solves a Real Business Problem?
I help startups, founders, and businesses build practical AI applications, from intelligent document processing to custom LLM-powered products.
I'm an AI/ML Engineer with experience in Generative AI, RAG, agentic workflows, NLP, and production AI systems. I focus on building solutions that go beyond a basic chatbot or AI demo.
Here's how I can help:
AI Document Intelligence & Automation
Extract structured information from PDFs, invoices, contracts, and business documents.
Automate document-heavy workflows and reduce repetitive manual work.
Build AI pipelines for document understanding, classification, and data extraction.
RAG Systems & AI Knowledge Assistants
Build chatbots that answer questions using your documents and internal knowledge base.
Implement embeddings, semantic search, retrieval optimization, and source-grounded responses.
Create internal knowledge assistants that help teams find information faster.
Custom AI Applications & Agentic Workflows
Develop LLM-powered applications and AI features.
Build multi-step AI workflows with tool integration and structured outputs.
Integrate AI models into Python backends and existing products using FastAPI and APIs.
How I approach projects
I start by understanding the business problem, then define the solution, build the core workflow, and validate it against real use cases. The goal is to deliver something useful, maintainable, and ready for the next stage of development.
One of my projects is Knowra, an enterprise knowledge assistant focused on document intelligence and retrieval-based AI.
If you're building an AI product, automating a business process, or exploring how LLMs could improve your existing software, I'd be happy to discuss your project.
Have an idea? Send me a message on Contra, and let's discuss what we can build.
Portfolio: https://portfolio-6lg.pages.dev/
Github: https://github.com/Ajeeetsingh
0
66
Metaforms - Tabulation Engine
I was the sole engineer on Metaforms' Tabulation Engine, owning the product, architecture, implementation and client demos. I built a Python compute layer over DuckDB that lets AI agents create crosstabs from uploaded specifications, with statistical testing, data weighting and confidence intervals. The system was benchmarked against WinCross and Q/Displayr.
0
15
Built a full ecommerce fashion storefront for AL Creations (abaya, hijab & custom wear) โ homepage, shop filters, product pages, cart/checkout, WhatsApp ordering, and a client admin to manage products & stock.
Designed for real small businesses who want an Amazon-style shopping experience without Shopify complexity.
Stack: HTML, CSS, JavaScript, React-ready UI patterns, Supabase, Netlify.
Live demo: https://al-creations-store-preview.netlify.app
I help brands launch clean, mobile-friendly websites that convert visitors into orders.
0
74
Project Title:- Custom Enterprise RAG Assistant | Chat with Documents & PDFs
Project Overview: DocuMind is an enterprise-grade Retrieval-Augmented Generation (RAG) assistant designed to eliminate manual data extraction and prevent AI hallucinations. It converts static corporate files (PDFs, reports, resumes, contracts) into an interactive, grounded knowledge engine that provides answers backed by direct page-level citations.
The Problem Solved: * Manual Data Hunting: Eliminates hours spent reading through dense, complex documentation.
AI Hallucinations: Constrains LLM outputs strictly to uploaded context, ensuring reliable, factual data.
Lack of Auditability: Provides exact file names and page references for compliance and verification.
Key Technical Features: * Dynamic Indexing: Fast chunking and local vector embedding using Hugging Face models (all-MiniLM-L6-v2).
High-Accuracy Vector Search: ChromaDB integration for persistent vector storage and low-latency similarity retrieval.
Decoupled Architecture: Asynchronous FastAPI backend paired with a clean, responsive Streamlit chat frontend.
Contextual Synthesis: Powered by Google Gemini (gemini-2.5-flash) for cost-effective inference.
Tech Stack: Python, FastAPI, LangChain, ChromaDB, Hugging Face, Google Gemini, Streamlit.
0
60
LookInsight AI : AI Customer Support Automation System
Full-stack AI system that automates customer support workflows.
What it does:
โ Classifies customer intent (complaint, inquiry, refund, support) โ Detects urgency and routes to the right team
โ Looks up customer history via knowledge graph
โ Generates personalized response drafts in seconds
Key features: โข Real-time processing pipeline with live visualization
โข Multi-channel support (Email, Twitter, Slack, WhatsApp)
โข Business rules engine for tier-based treatment (VIP, Premium, Regular)
โข Neo4j knowledge graph for customer context
Tech stack: Python, FastAPI, OpenAI GPT-4, Neo4j, Next.js, TypeScript
Live demo: https://lookinsight.ai
Skills: Python AI OpenAI FastAPI Next.js Customer Support Automation
Link: https://lookinsight.ai
1
0
78
Vision-Cart V2: Neural Intelligence Platform
Developed a high-end Visual Auditing dashboard for retail and industrial quality control. Unlike basic classifiers, this system performs Deep Feature Extractionโcalculating metrics like caffeine content (mg), volume (ml), and shelf fill-levels.
Edge-AI Mastery: Optimized to run Ollama (Moondream) locally on restricted hardware (2GB VRAM), ensuring total data privacy.
Industrial UI: Built a 'Nuvion-Black' interface with real-time scanning animations and a horizontal Bento-grid for system vitals.
Business Logic: Integrated a 'Neural Suggestion' engine to turn visual scans into automated inventory and QC decisions.
Tags: #ComputerVision #EdgeAI #React #FastAPI #MachineLearning
0
74
Most sales teams don't struggle because of a lack of leads.
They struggle because finding, qualifying, researching, and reaching out to prospects is still largely manual.
HunterOS is an AI-powered sales prospecting and outreach platform designed to automate the top of the sales funnel.
It continuously identifies potential customers, gathers relevant business intelligence, qualifies opportunities, and prepares personalized outreach at scale.
Instead of spending hours searching for prospects, sales teams receive a pipeline of high-potential opportunities ready for engagement.
The result:
โ Faster prospect discovery
โ Reduced manual research
โ Higher sales productivity
โ Consistent pipeline generation
โ More time spent closing deals
HunterOS transforms prospecting from a repetitive task into an intelligent, automated process.
#SalesAutomation #LeadGeneration #AIAutomation #BusinessGrowth #SalesOps #HunterOS
0
75
Project Goal:
Designed and developed a full-stack, AI-powered e-waste management platform to promote responsible electronic waste disposal while enabling value recovery, recycling discovery, and convenient pickup services.
โข Built RESTful backend APIs using FastAPI / Node.js for authentication, classification workflows, facility management, pickup scheduling, and marketplace operations
โข Integrated an AI image classification model to categorize e-waste into recyclable, reusable, or hazardous types
โข Implemented a facility locator with map integration to identify nearby certified recycling centers
โข Developed a rule-based value estimation engine for resale and scrap pricing
โข Created a slot-based pickup scheduling system with confirmation flow
โข Integrated Razorpay payment gateway for secure online transactions
0
153
REPOLENS
a deployed ai code-intelligence platform with 27 pre-indexed repositories and 60,374 real source chunks. classic rag and a tool-using google adk agent answer codebase questions with file-and-line citations, backed by fastapi, postgresql/pgvector, gemini, neon, docker, and 30 automated tests.
1
35
๐๐ฒ๐ฎ๐ฑ๐น๐ถ๐ป๐ฒ: Most Voice AI agents still feel like glorified walkie-talkies. So I built one that actually listens like a human.
Most voice bots today follow a rigid, linear loop: You speak โ Wait โ Bot speaks. If you dare interrupt them mid-sentence? They either completely ignore you or their context buffer gets totally mangled.
๐ง๐ผ ๐๐ผ๐น๐๐ฒ ๐๐ต๐ถ๐, I engineered Astra Duplex Agent โ a sub-50ms, ultra-low latency, full-duplex conversational Voice AI built to handle natural human interruptions seamlessly.
Here is what went into building the architecture under the hood:
True Full-Duplex WebSockets: Built on FastAPI, allowing real-time bi-directional audio streaming instead of turn-based HTTP requests.
Edge Voice Activity Detection (VAD): Deployed Silero VAD using ONNX Runtime to catch user barge-ins at the edge with near-zero overhead.
LangGraph State Machine: Instead of a simple monolithic script, the conversation logic is modeled as a state graph, giving precise control over execution flow.
Partial State Tracking & Memory Persistence: This was the trickiest part. When an interruption happens, Astra doesn't just stop audio playback โ an async kill-switch halts Groq's LLaMA-3.1 mid-token, calculates the exact partial sentence actually spoken by the TTS, and saves only that partial context to an Upstash Redis buffer.
Sub-50ms Inference & Streaming: Leveraged Groq LPU (LLaMA-3.1 + Whisper Large v3) paired with ElevenLabs Turbo v2.5 streaming.
The result? An agent that you can interrupt mid-thought, change topics with on the fly, and ask follow-up questions without it losing historical context.
Containerized with Docker and live on Render & Vercel.
Check out the full architecture & live demo below!
https://lnkd.in/eC6yak3i
(https://lnkd.in/eC6yak3i)https://lnkd.in/eN38HGZD
(https://lnkd.in/eN38HGZD)#VoiceAI (https://www.linkedin.com/search/results/all/?keywords=%23voiceai&origin=HASH_TAG_FROM_FEED) #GenerativeAI (https://www.linkedin.com/search/results/all/?keywords=%23generativeai&origin=HASH_TAG_FROM_FEED) #LangGraph (https://www.linkedin.com/search/results/all/?keywords=%23langgraph&origin=HASH_TAG_FROM_FEED) #SystemDesign (https://www.linkedin.com/search/results/all/?keywords=%23systemdesign&origin=HASH_TAG_FROM_FEED) #FastAPI (https://www.linkedin.com/search/results/all/?keywords=%23fastapi&origin=HASH_TAG_FROM_FEED) #MachineLearning (https://www.linkedin.com/search/results/all/?keywords=%23machinelearning&origin=HASH_TAG_FROM_FEED) #WebSockets (https://www.linkedin.com/search/results/all/?keywords=%23websockets&origin=HASH_TAG_FROM_FEED) #Python (https://www.linkedin.com/search/results/all/?keywords=%23python&origin=HASH_TAG_FROM_FEED) #AIENGINEERING (https://www.linkedin.com/search/results/all/?keywords=%23aiengineering&origin=HASH_TAG_FROM_FEED)
1
76
