Initialising Cosmic Interface
✦ Computer Science & Engineering  ·  B.Tech  ·  Batch 2027  ·  PSIT Kanpur ✦

MAN-RA

GenAI & Full-Stack AI Developer

_

Scroll

About Me

Hello, I'm Man-Ra

A B.Tech CSE student (Batch 2027) at PSIT Kanpur with a strong AI/ML and GenAI foundation and a habit of shipping end-to-end systems — from predicting antibiotic resistance directly from raw genomic sequences, to ParaFlow AI, an LLM-powered text orchestration platform, to DevCollab, a freelancer-client collaboration platform I'm building right now.

My stack spans Python, FastAPI, Next.js, PostgreSQL, Supabase, Redis and Docker, with LLM integrations through the OpenAI API and Gemini. I care about the full path from model to product — REST APIs, authentication, caching, payments and deployment — not just the notebook.

Preparing for campus placements 2027 and open to SDE and AI/ML roles where I can build things that matter.

3Personal Projects
2027B.Tech CSE Batch
4Certifications
GenAICore Focus

Skills

Languages & Core

Python90%
HTML / CSS78%
JavaScript / TypeScript65%
SQL70%

AI / GenAI / Agentic

GenAI & LLM Applications82%
Agentic AI Systems78%
ML / DL Fundamentals (CNN, Transformers)75%

Tools & Platforms

VS CodeJupyterGoogle ColabGitGitHub FastAPINext.jsReactSupabase PostgreSQLRedisDockerWebRTCRazorpay OpenAI APIGemini / Google ADKn8n Vibe CodingAI AgentsAutomation

Core Subjects

DBMSOperating SystemsComputer NetworksSoftware EngineeringMachine Learning

Education

🎓

B.Tech — Computer Science Engineering

Pranveer Singh Institute of Technology, Kanpur 2024 – Present
📘

Diploma — Computer Science Engineering

Kanpur Institute of Technology, Kanpur 2022 – 2024
🏫

Class XII

Kanpur City Inter College, Kanpur 2021 – 2022
🏫

Class X

Kanpur City Inter College, Kanpur 2019 – 2020

Certifications

🧠

BCG GenAI Job Simulation

Boston Consulting Group × Forage Completed · February 2026
📋

Agile Product Management

Great Learning Completed
🤖

AI Automation using N8N

CampusX Completed
🐍

Python Programming Fundamentals

Infosys Springboard Completed · 2025

Resume

Every project, every model, every pipeline — condensed into one document.

↓ Download Resume (PDF)

Contact

Let's Build Something

SDE and AI/ML roles, collaborations, project ideas, or just a conversation about building AI-powered products — reach out. I'm always building, always curious.

Personal Project · Bioinformatics & Deep Learning

Antibiotic Resistance Prediction via Pathogen DNA

StatusCompleted · Ready to Deploy
DomainBioinformatics / CNN
YearAug – Dec 2025
RoleSolo Developer
PythonBioPythonTensorFlowCNNBLT ArchitectureTransformer
▶  Project Demo
▶
Paste your YouTube / Loom embed ID below in the code Find: YOUR_P1_VIDEO_ID and replace

Overview

Antibiotic resistance kills 1.3 million people annually. Traditional culture-based diagnostics take 24–72 hours — time that costs lives in critical infections. Genora builds a BLT-CNN pipeline that reads raw 300-nucleotide pathogen DNA sequences and predicts resistance patterns in milliseconds, with real-world potential for rapid screening in clinical microbiology.

The Problem

Without knowing resistance patterns, clinicians prescribe broad-spectrum antibiotics by default — accelerating global resistance. The core question: can we extract resistance signals directly from DNA sequences before any culture result arrives?

The Architecture

Raw sequences from the CARD database are segmented by an entropy-guided BLT patcher. A 1D CNN captures local sequence motifs while a 4-layer, 8-head Transformer encoder models longer-range context. Both are fused through an MLP into a multi-label sigmoid classifier (~944K parameters) that outputs one resistance probability per antibiotic class.

Key Features

Learns directly from raw DNA with no manual feature engineering. Entropy-guided BLT patching. Hybrid 1D CNN + Transformer encoder. Multi-label resistance prediction across Methicillin, Ciprofloxacin and Vancomycin.

Results

Trained on the CARD database, the model reaches 93.34% accuracy, 98.55% AUC, 94.18% F1, 95.57% precision and 93.35% recall across the three antibiotic classes.

Personal Project · GenAI & Full-Stack SaaS

ParaFlow AI: End-to-End IntelliScribe Autonomous Text Orchestration System

StatusCompleted
DomainGenAI / SaaS
YearApr 2026 – Sep 2026
RoleDeveloper
PythonFastAPINext.jsOpenAI APIPrompt EngineeringREST APIsPostgreSQL (Supabase)RedisDocker
▶  Project Demo
▶
Add your demo video — replace YOUR_P2_VIDEO_ID

Overview

ParaFlow AI is an end-to-end AI writing platform. It accepts long prompts (5,000+ characters) and turns them into polished content through writing, rewriting, summarization, translation and grammar correction — all from one production-ready SaaS.

What It Does

Users submit a prompt, choose a task, and get AI-generated content back in 2–10 seconds. The platform maintains persistent user history and tracks usage, so past work stays available and consumption stays visible.

Architecture

A Next.js frontend talks to a FastAPI backend exposing 15+ REST APIs behind JWT authentication. Prompt-engineered LLM pipelines call the OpenAI API, Redis caches repeated work, PostgreSQL (Supabase) stores history and usage, and the whole system is containerized with Docker for secure cloud deployment.

Results

Generates high-quality AI content within 2–10 seconds on 5,000+ character prompts, backed by a scalable, production-ready SaaS architecture.

Personal Project · Full-Stack · Marketplace · AI Agents

DevCollab — Freelancer-Client Collaboration Platform

StatusCurrently Building 🔨
DomainFull-Stack / Marketplace
YearSep 2026 – Present
RoleDeveloper
Next.jsTypeScriptTailwind CSSFastAPISupabaseWebRTCRazorpayGoogle ADKGemini
▶  Project Demo
▶
Demo coming soon — currently building

Overview

DevCollab is a collaboration platform where freelancers and clients find each other, break projects into milestones, chat, track tasks and jump on video calls — with payments held against milestones until the work is approved.

What It Does

A marketplace and dashboard for projects. Milestone-based payments through Razorpay with webhook-verified payment states. Real-time chat and presence. Task tracking per project. Peer-to-peer 1-to-1 video and audio calls over WebRTC. Google ADK agents powered by Gemini for AI-assisted project workflows.

Architecture

A Next.js + TypeScript + Tailwind frontend and a FastAPI backend that verifies Supabase-issued JWTs. Supabase provides PostgreSQL with row-level security, Auth, Storage and Realtime — Realtime channels also carry the WebRTC signaling.

Current Status

In active development (Sep 2026 – Present), built in phases. The demo and repository will be linked here once the core flow is live.