PK.
AI / ML engineerMumbai, Maharashtra

Prathmesh
Kangane.

I build practical AI systems for research, interview practice, and speech understanding.

Open to entry-level rolesPune · Bangalore · Hyderabad · Mumbai
Explore my work View resume Download resume
Laptop showing a research-paper search interface beside academic papers and a notebook
Current focusRetrieval · grounding · multimodal AIBUILD / EVALUATE / REFINE
3
end-to-end AI projects
AI / ML
engineering graduate
Research
published · April 2026
Immediate
joining availability
01Selected projects

Built, tested, measured.

DEC 2025 — NOV 2026
PROJECT 01Dec 2025 — Nov 2026

AI Research Assistant

A citation-grounded research companion that turns a library of papers into precise, evaluable answers.

  • Built a full-stack search application for 10,000 ArXiv papers, with a FAISS index, authentication, and Docker Compose.
  • Measured retrieval on 141 LLM-generated queries; BGE-small achieved 0.87 keyword Recall@1 and 0.90 MRR.
  • Grounded Gemini answers in numbered paper citations; automated a 30-test suite with GitHub Actions CI.
PythonFlaskMySQLBGE-smallFAISSGemini APIDockerpytest
View GitHub
Retrieval evaluation
MiniLMBGE-smallMeasured results · Recall@k and MRR
Recall@10.00–1.00
0.770.87
00.51
MRR0.00–1.00
0.820.90
00.51
PROJECT 02Feb 2026 — Oct 2026

AI-Powered Mock Interview Platform

A multimodal practice interview that tailors questions to a candidate and turns every session into a performance report.

  • Generates personalized interview questions from a resume and job description, then labels each question type with fine-tuned DistilBERT.
  • Pairs a live interview room with real-time speech-to-text and facial emotion tracking.
  • Combines fluency, STAR structure, confidence trends, and Gemini-based content evaluation in a hybrid-score PDF report.
PythonFastAPIGemini APIDistilBERTHugging FacePyTorchspaCy
View GitHub
A structured interview loop
  1. STEP 01ContextResume + job description
  2. STEP 02GenerateGemini writes questions
  3. STEP 03ClassifyDistilBERT labels type
  4. STEP 04PracticeLive room · speech-to-text · facial emotion
  5. STEP 05ReviewDownloadable PDF report
PROJECT 03Nov 2025 — Jun 2026

Speech Emotion Recognition

A real-time speech emotion classifier that pairs a stacked LSTM with an interactive, emotion-aware app.

  • Trains on RAVDESS and TESS features: MFCC, chroma, and log-mel spectrograms.
  • Identifies seven emotion classes and runs in a real-time Streamlit application.
  • Reached 0.81 test accuracy and 0.81 macro F1 on a held-out test set; includes an emotion-aware chatbot.
PythonTensorFlowLibrosaMFCCStacked LSTMStreamlit
View GitHub
Model snapshot
Test accuracy0.81
Macro F10.81
Held-out test set7 emotion classes
02Experience

Learning by doing.

ALL INTERNSHIPS · VIRTUAL

Power BI & Data Analytics Intern

Microsoft Elevate · via AICTE

Feb 2026 — Mar 2026Virtual internship
  • Completed Microsoft Learn paths covering Power BI, DAX, and Microsoft Fabric.
  • Built a Power BI sales dashboard from the UCI Online Retail dataset, cleaning data with Power Query and defining DAX measures.

Computer Vision Intern

Infosys Springboard

Feb 2026 — Apr 2026Virtual internship
  • Built a Streamlit image-stylization app with ten OpenCV and Pillow filters, user sign-in, personal history, and SQLite storage.
  • Deployed the application on Streamlit Community Cloud.

AI & Cloud Computing Intern

Edunet Foundation · in collaboration with IBM SkillsBuild & AICTE

Jul 2025 — Aug 2025Virtual internship
  • Compared crop-recommendation pipelines in IBM watsonx.ai AutoAI and deployed the top-ranked pipeline as an online endpoint.
  • Configured a LangGraph ReAct agent with search tools in watsonx.ai Agent Lab.
03Research & recognition
Published researchAPRIL 2026

Multimodal AI System for Automated Interview Analysis

International Journal of Engineering Development and Research (IJEDR)

Vol. 14 · Issue 2Paper ID IJEDR26021594 co-authors
View publication certificate

Certifications

6 COMPLETIONS
04About

Curious about the whole system.

Computer Science (AI & ML) engineering graduate from Saraswati College of Engineering, Kharghar — University of Mumbai. I work across retrieval-augmented generation, NLP, speech emotion recognition, and data analytics.

Bachelor of Engineering — Computer Science (AI & ML)

Saraswati College of Engineering, Kharghar · University of Mumbai

JUNE 2022 — MAY 2026 · CGPI 7.7

Technical toolkit

LanguagesPython · Java · SQL
FrameworksFastAPI · Flask · Streamlit · TensorFlow · PyTorch
AI / MLNLP · RAG · Transformers · Semantic search · Speech emotion recognition
LibrariesScikit-learn · Pandas · NumPy · Librosa · OpenCV
Platforms & toolsGemini API · FAISS · LangGraph · Docker · GitHub Actions · Power BI

Also studied: database management, software design, data structures & algorithms, statistics, machine learning, and deep learning.

05Open to opportunities

Let’s make
something useful.

Entry-level AI/ML EngineerGenAI EngineerMachine Learning EngineerData Scientist
PREFERREDPuneBangaloreHyderabadMumbai
Available for immediate joining
Get in touch

prathmesh.device23@gmail.com

Email me
LinkedIn GitHub