I Build AI Systems That Replace Manual Work.
I help startups automate operations using Agentic AI, LLMs, RAG systems, AI Voice Agents and workflow automation.
0+
AI Systems Shipped
0+
Automation Hours Saved / mo
0
Published Research
0+
Client Domains
I don't sell chatbots. I solve operational problems.
Every engagement starts with a business bottleneck, not a model. Here's what I actually get hired to fix.
Reduce Repetitive Work
Offload manual, repetitive operations to autonomous agents that run 24/7 without supervision.
Automate Customer Support
Deploy AI agents that resolve tickets, answer questions, and escalate intelligently — not scripted chatbots.
Internal AI Assistants
Give your team an assistant that knows your company's docs, tools, and workflows inside out.
AI-Powered Knowledge Bases
Turn scattered documents into a searchable, cited, hallucination-checked RAG system.
WhatsApp Automation
Meta WhatsApp-integrated agents that qualify leads, book calls, and support customers at scale.
Voice Calling Agents
Human-like voice agents that call, qualify, and log leads automatically — no human dialing required.
Sales Outreach Automation
Personalized, duplicate-safe cold outreach generated and sent on autopilot.
Document Intelligence
Extract, summarize, and reason over contracts, reports, and case files with retrieval-grounded accuracy.
Business Process Automation
Connect your tools end-to-end with n8n and custom APIs — data flows without human handoffs.
How I work
Systems over demos.
- Ships production code, not notebooks
- Evaluation & monitoring on every system
- Fast iteration without cutting corners
I got into AI to build things that actually ship.
I'm an AI/ML Engineering graduate who spent more time in production logs than in lecture slides. Somewhere between training my first CNN and debugging my first agent that refused to stop calling the same API in a loop, I realized what I actually care about isn't the model — it's whether the system survives contact with real users and real edge cases.
That shift is what led to NyayaRAG, my published research on KisanMitra, and the autonomous agents, voice systems, and automation pipelines I now build for clients. I don't ship demos — I ship systems with evaluation, monitoring, and error handling baked in, because that's the difference between an AI toy and an AI system a business can depend on.
If you're a founder or ops lead staring at a process that shouldn't need a human anymore, that's exactly the problem I like solving.
Systems I've shipped, not just prototyped.
Four production-oriented builds — each with the problem, the architecture, and the business outcome.
NyayaRAG
A production-grade legal AI assistant with evaluated, hallucination-checked retrieval.
Problem
Legal teams drown in case law and contracts. Generic LLM chat hallucinates citations — an unacceptable risk in legal work.
Solution
A retrieval-augmented legal assistant that grounds every answer in retrieved source documents, re-ranks for precision, and continuously evaluates itself for hallucination and recall before answers ship.
Key features
Business impact
- Grounded answers with source citations instead of free-form hallucination
- Automated eval suite catches regressions before deploy
- Sub-second retrieval on a re-ranked hybrid index
Architecture flow
Ingestion pipeline chunks & embeds legal documents via LlamaIndex
Hybrid retrieval — BM25 keyword search + dense vector search over Chroma
Cross-encoder re-ranking narrows candidates to the most relevant passages
FastAPI backend serves grounded, cited responses
Ragas + DeepEval run automated context recall & hallucination scoring in CI
Dockerized services deployed via GitHub Actions
Latency & token-cost monitoring on every request
AI Voice Calling Agent
A voice AI agent that calls, qualifies, and books leads — fully automated.
Problem
Sales teams lose hours dialing unqualified leads, and outcomes never make it back into a clean spreadsheet.
Solution
An n8n-orchestrated voice pipeline where Twilio places calls, ElevenLabs delivers a natural conversational voice, and every outcome — qualified, callback, not interested — is logged straight into Google Sheets with scheduling handled automatically.
Key features
Business impact
- Eliminates manual dialing entirely
- Every call outcome captured with zero manual data entry
- Scales outbound calling without adding headcount
Architecture flow
n8n workflow triggers outbound calls from a lead queue
Twilio handles telephony & call routing
ElevenLabs generates natural, low-latency conversational voice
Conversation AI layer handles objections & qualification logic in real time
Call outcomes and transcripts logged to Google Sheets
Automated scheduling for qualified leads
Agentic Cold Outreach
Autonomous, personalized cold email generation with built-in duplicate prevention.
Problem
Cold outreach at scale is either generic (low reply rate) or hand-written (doesn't scale). Most tools also re-email the same lead twice.
Solution
An autonomous agent that reads lead data from Google Sheets, researches context, drafts a genuinely personalized email per lead, sends via the Gmail API, and tracks who's already been contacted to guarantee zero duplicates.
Key features
Business impact
- Higher reply rates from genuinely personalized copy
- Zero duplicate outreach to the same lead
- Outreach volume scales without a growth team
Architecture flow
Lead source: Google Sheets as the single source of truth
Agent enriches each lead with contextual research
LLM drafts a personalized email per lead — not a template with merge fields
Duplicate-prevention layer checks send history before dispatch
Gmail API sends and logs delivery status back to the sheet
KisanMitra
An AI agriculture platform for crop recommendation and plant disease detection.
Problem
Smallholder farmers often lack fast access to agronomic expertise — the wrong crop choice or a missed disease can cost a whole season.
Solution
A machine learning platform that recommends the optimal crop from soil & weather data and detects plant diseases from leaf images using a trained CNN — served through a simple Flask web app.
Key features
Business impact
- Published as peer-reviewed research at FutureTech 2025
- Gives farmers a data-backed second opinion in seconds
- Demonstrates end-to-end ML: data → model → deployed app
Architecture flow
Crop recommendation model trained with Scikit-learn on soil & climate data
Disease detection CNN built and trained in TensorFlow
Live weather data pulled from a public Weather API
Flask backend serves both models through a unified interface
TransitGo India
A production iOS ride-booking platform — live on the App Store.
Problem
Riders needed a booking experience that felt safe, not just functional — with live trip visibility and a way to signal for help instantly if something went wrong.
Solution
A full ride-booking platform shipped to the App Store: instant booking, live driver tracking on a map, secure in-app payments, multiple vehicle classes, and a built-in emergency SOS. Built and released as part of a cross-functional engineering team at Shankh Technologies.
Key features
Business impact
- Live on the iOS App Store with a public release
- Shipped end-to-end as part of a production engineering team
- Safety-first design: SOS and live tracking built into the core ride flow
Architecture flow
Mobile client handles booking flow, map rendering, and live trip state
Real-time driver location streamed to the rider during an active trip
Backend APIs coordinate matching, trip lifecycle, and ride history
Secure payment gateway integration for in-app transactions
Emergency SOS triggers an immediate alert path from inside an active ride
Shipped through the full App Store review and release pipeline
A stack built for production, not demos.
Every tool here has shipped in a real client system — not just a side project.
Programming
Generative AI & LLMs
ML & Deep Learning
Web Development
Mobile Development
Automation
Databases
Deployment & Tools
Engineering judgment, not just prompt-writing.
Anyone can call an LLM API. Here's what actually makes a system worth paying for.
Production AI Systems
I ship evaluated, monitored systems — not notebook demos.
Business-First Mindset
Every system is designed around a business metric, not just a model score.
Fast Development
Lean toward working prototypes in days, not sprints of slideware.
Clean Architecture
Modular, testable systems your team can extend after I'm gone.
Scalable Solutions
Built to handle real traffic — async, containerized, monitored.
Automation Expert
I connect your existing tools instead of asking you to replace them.
LLM Specialist
Deep, hands-on experience across OpenAI, Claude, Gemini, and open models.
GenAI Product Development
End-to-end GenAI delivery — web app, mobile app, and deployment — not just the model layer.
KisanMitra: AI-Driven Smart Agriculture System
A peer-reviewed paper presenting an end-to-end AI platform for crop recommendation and plant disease detection, combining classical ML and deep learning to give farmers real-time, data-backed agronomic guidance.
Read the publicationFrom classroom to production systems.
Started B.E. in AI & Machine Learning
Began formal AI/ML engineering education, building a foundation in statistics, deep learning, and systems thinking.
First production ML systems
Built and shipped early crop-recommendation and computer-vision projects — the seeds of KisanMitra.
Published KisanMitra research
Co-authored and published "KisanMitra: AI-Driven Smart Agriculture System" at FutureTech 2025.
Software Development Intern — Shankh Technologies
Built and shipped production software on real client projects as part of a cross-functional engineering team, including the TransitGo India iOS app released on the App Store.
Agentic AI systems — NyayaRAG, Voice Agents, Outreach Automation
Shifted focus to agentic architectures: evaluated RAG pipelines, voice AI agents, and autonomous outreach systems.
Freelance AI Automation Engineer
Started building production AI automation systems for startups and businesses — from WhatsApp agents to workflow automation.
Live from GitHub.
Pulled straight from the API — real repos, real activity, no curated screenshots.
What it's like to work together.
“Tousif shipped an AI agent that replaced three hours of daily manual work in under two weeks. The kind of engineer who actually understands the business problem.”
Founder
SaaS Startup
“The RAG system he built for us is the first AI tool our legal team actually trusts — because it cites its sources and doesn't hallucinate.”
Operations Lead
Law Firm
“Our WhatsApp support agent now handles most tier-1 tickets automatically. Setup was fast, and the handoff logic to humans is seamless.”
Head of Support
Ecommerce Brand
Pick a problem. I'll build the system.
Fixed-scope engagements or ongoing partnership — whatever gets your AI system shipped.
AI Chatbots
Context-aware chatbots that actually understand your product and customers.
AI Agents
Autonomous agents that reason, plan, and take action across your tools.
RAG Systems
Retrieval-augmented systems grounded in your documents, evaluated for accuracy.
Voice Agents
Natural-sounding AI voice agents for calls, qualification, and support.
Workflow Automation
End-to-end automation of business processes with n8n and custom APIs.
WhatsApp Automation
Meta WhatsApp-integrated AI agents for sales, support, and lead qualification.
LLM Integrations
Ship OpenAI, Claude, and Gemini deep into your product — done right.
GenAI Web & Mobile Apps
AI-powered web and mobile products built and shipped to production — including the App Store.
Custom AI Solutions
A specific business problem, solved with the right AI system for the job.
Let's Build Something Amazing
Have a process that shouldn't need a human anymore? Tell me about it — I reply to every serious inquiry within 24 hours.
