Skip to content
View iam-teju's full-sized avatar

Block or report iam-teju

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
iam-teju/README.md

Hi, I'm Tejas Manoj

Data & ML Engineer with 2.5 years of experience building data pipelines and ML systems. B.Tech in Computer Science from VIT Vellore. Published researcher in NLP — my work on predicting Indian film success using subtitle-derived document vectors was published in the International Journal of Image and Graphics (World Scientific, 2023).

Currently building AI products across computer vision, deep learning, and NLP.

What I work with

Languages & Frameworks: Python, SQL, FastAPI, Flask, Streamlit
ML & Deep Learning: PyTorch, TensorFlow, Scikit-learn, XGBoost, Hugging Face Transformers
Computer Vision: MediaPipe, OpenCV, dlib
NLP & LLMs: LangChain, Sentence Transformers, Doc2Vec, Claude API
Data & Infrastructure: Snowflake, dbt, ChromaDB, Supabase, PostgreSQL
Tools: Git, Docker, CUDA/cuDNN, Jupyter

Selected Projects

Project What it does Stack
FormCheck AI Real-time AI gym coach — detects exercise form via pose estimation and gives live voice feedback MediaPipe, OpenCV, Streamlit, pyttsx3
ClassSense AI-powered attendance system using face recognition (dlib + SVM) and voice authentication (Resemblyzer) dlib, SVM, Resemblyzer, Streamlit, Supabase
StyleCast Neural style transfer web app using AdaIN — trained a custom decoder on 40K+ images (MS-COCO & WikiArt), deployed with Flask on Render PyTorch, VGG-19, Flask, AdaIN
Film Success Prediction NLP pipeline predicting movie profitability from subtitles using Doc2Vec embeddings (F1: 0.77) — published paper Doc2Vec, XGBoost, AdaBoost, Scikit-learn

Research

Early Success Prediction of Indian Movies using Subtitles: A Document Vector Approach
International Journal of Image and Graphics, World Scientific, Vol. 23, No. 4, 2023
Built a two-stage ML pipeline (regression → classification) using custom subtitle vectors across 2700+ Indian films in five languages. Achieved F1-score of 0.77 and Cohen's Kappa of 0.48.

Links

LinkedIn Email

Pinned Loading

  1. ClassSense ClassSense Public

    AI-powered attendance system using Face Recognition (dlib + SVM) and Voice Authentication (Resemblyzer). Built with Streamlit and Supabase.

    Python

  2. StyleCast StyleCast Public

    Real-time Neural Style Transfer web app using Adaptive Instance Normalization (AdaIN). Built with PyTorch and Flask.

    HTML

  3. FormCheck-AI FormCheck-AI Public

    Real-time AI gym coach using computer vision for form detection and live voice coaching

    Python

  4. Early-Success-Prediction-Of-Indian-Movies Early-Success-Prediction-Of-Indian-Movies Public

    Two-stage ML pipeline predicting Indian film success from subtitle embeddings (Doc2Vec). Published in IJIG, World Scientific 2023.

    Python