KKaran ShihireGet in touch
Pune, Maharashtra, India

Karan Shihire

AI Engineer building semantic search, computer vision, and LLM-powered products.

I turn machine learning ideas into practical, deployable applications with Python, FastAPI, vector search, and modern frontend tooling.

View ProjectsDownload ResumeContact Me
AI workflow illustration showing data ingestion, vector search, API services, and frontend delivery
About

Building AI Products from Idea to Deployment

I enjoy solving real engineering problems by building complete AI applications instead of isolated machine learning models.

My work spans semantic search, recommendation systems, computer vision, retrieval pipelines, REST APIs, and deployment.

I focus on writing maintainable backend services, designing modular architectures, and integrating AI models into usable products.

Applied AI
Primary Focus
FastAPI
Backend
React
Frontend
Pune
Location
GitHub activity

Open-source and project activity

A quick snapshot of my public GitHub profile and recent development activity.

Repositories
10
Stars
4
Primary Languages
Python
Latest Commit
FaceTrack • July 4, 2026

Recent contribution pattern

Public profile metrics verified from GitHub on July 4, 2026.

View GitHub
Technology stack

Technology Stack

Technologies I regularly use while building AI and backend applications.

Languages

3
PythonSQLJava

AI / ML

15
Machine LearningDeep LearningComputer VisionLarge Language ModelsPrompt EngineeringSemantic SearchEmbeddingsRecommendation SystemsVector DatabasesFAISSOpenCVInsightFaceNumPyPandasScikit-learn

Backend

2
FastAPIREST APIs

Frontend

3
ReactTypeScriptTailwind CSS

Databases

1
PostgreSQL

Tools

6
GitGitHubVS CodeCursorRenderNetlify
Featured projects

Featured Projects

Production-style AI applications demonstrating semantic search, computer vision, machine learning, and backend engineering.

01 · Project

CareerLens

Semantic resume intelligence and job recommendation engine

Problem

Finding relevant jobs through keyword search often misses semantic meaning across skills, roles, and resume context.

Solution

Built a recommendation engine using transformer embeddings, FAISS vector search, hybrid ranking, and LLM-generated resume summaries to match resumes with relevant opportunities.

Impact

Designed a modular retrieval pipeline that can surface relevant jobs quickly while also highlighting skill gaps and improving the interpretability of recommendations.

Architecture
  1. 1Resume ingestion
  2. 2Parsing
  3. 3Embeddings
  4. 4Vector retrieval
  5. 5Reranking
  6. 6Recommendation output
Key Features
  • Resume parsing from PDF files
  • Transformer-based semantic embeddings
  • FAISS-powered vector retrieval
  • Hybrid retrieval and reranking pipeline
  • LLM-generated resume summaries
  • Skill gap analysis
  • FastAPI backend
  • React and TypeScript frontend
Tech Stack
PythonFastAPISentence TransformersFAISSOpenAI APIReactTypeScript
DeploymentGitHub
02 · Project

FaceTrack – Real-Time Face Recognition Platform

Attendance tracking with live recognition and vector similarity search

Problem

Attendance systems based on manual entry or basic matching struggle with accuracy, speed, and usability in live multi-person scenarios.

Solution

Developed a real-time face recognition platform using InsightFace embeddings, OpenCV, FAISS similarity search, FastAPI services, and a React dashboard for enrollment and tracking.

Impact

Combined enrollment, recognition, and attendance logging into a usable workflow suited for real-time operation and backend-driven integration.

Architecture
  1. 1Enrollment
  2. 2Video stream
  3. 3Face detection
  4. 4Embeddings
  5. 5Similarity search
  6. 6Attendance records
Key Features
  • Multi-angle face enrollment
  • Live webcam recognition
  • Deep face embeddings
  • FAISS similarity matching
  • Attendance logging
  • FastAPI backend
  • PostgreSQL integration
  • React dashboard
Tech Stack
PythonFastAPIOpenCVInsightFaceONNX RuntimeFAISSReactPostgreSQL
Source Code
03 · Project

NetWatch – Intelligent Network Intrusion Detection System

Live traffic inspection with anomaly detection and alerting

Problem

Suspicious network behavior is difficult to spot manually when traffic is continuous and event volume grows over time.

Solution

Built an intrusion detection workflow that captures packets with Scapy, extracts traffic features, and uses Isolation Forest to identify anomalous behavior through a Flask dashboard.

Impact

Turned raw packet streams into a monitoring interface that can surface unusual activity in real time and support quick operator review.

Architecture
  1. 1Packet capture
  2. 2Feature extraction
  3. 3Anomaly model
  4. 4Alerting
  5. 5Dashboard
Key Features
  • Live packet capture
  • Packet feature extraction
  • Isolation Forest anomaly detection
  • Scapy-based traffic inspection
  • Real-time monitoring
  • Alert generation
  • Flask dashboard
Tech Stack
PythonFlaskScapyScikit-learnIsolation Forest
Source Code
Engineering journey

Engineering Journey

A snapshot of the technical areas I've explored while building end-to-end AI applications.

  1. 01

    Semantic Search & Recommendation Systems

    Built retrieval pipelines using embeddings, FAISS, hybrid ranking, and LLM-assisted recommendations.

  2. 02

    Computer Vision

    Developed real-time face recognition systems using InsightFace, OpenCV, and vector similarity search.

  3. 03

    Backend Engineering

    Designed FastAPI services with modular architecture, API integrations, and scalable project organization.

  4. 04

    Deployment

    Deployed frontend and backend applications using Netlify and Render while managing production configurations.

  5. 05

    Continuous Learning

    Currently expanding into Retrieval-Augmented Generation (RAG), LLM evaluation, AI agents, and scalable inference systems.

Contact

Interested in Building AI Products Together?

I'm open to roles and collaborations focused on applied AI, backend engineering, and product-oriented software development.

Download resume

This opens your email client with the message details.