Aniket Deshpande

Aniket Deshpande

Data Scientist & Machine Learning Engineer

View Projects Download Resume
AWS Machine Learning Specialty

Scroll to see more ↓

Aniket Deshpande

Projects

End‑to‑end ML systems, RAG pipelines, and real‑time analytics — shipped with CI/CD and thoughtful evaluation.

LLM NLP App — API + UI on Hugging Face Spaces

LLM NLP App — API + UI (Spaces)

Modular FastAPI + Gradio stack delivering text summarization & rewrite with transformer models. Fully automated deploys to two Spaces via GitHub Actions.

FastAPIGradioTransformersCI/CD
Read More

What I built

  • Separate API and UI services with health checks, env validation, and smoke tests.
  • Automated CI/CD: push → build → deploy to Spaces; blue‑green style swap on success.
  • Fallback routing (local models → OpenRouter/OpenAI) with graceful degradation.

Links

Live API Live UI View Code
Stockio — forecasting & sentiment dashboard

Stockio — Forecast & Sentiment

Interactive stock dashboard with fast price forecasts, news sentiment, and sanity checks. Supports multi‑ticker comparisons.

StreamlityfinanceProphet/ARIMAPlotly
Read More

What I built

  • Cache‑aware pipelines for fetching quotes & generating short‑horizon forecasts.
  • Forecast sanity guardrails (bounds, trend checks) to curb over‑confident outputs.
  • Clean UX with historical chart, forecast table, and quick filters.

Links

Live Demo View Code
Minimal free-tier RAG QA over SEC filings

RAG‑QA (Free‑Tier, EDGAR 10‑K)

CPU‑only retrieval‑augmented QA using MiniLM + FAISS + extractive QA. Automates EDGAR ingestion for Risk/MD&A sections and returns citation‑grounded answers.

FAISSSentence‑TransformersEDGARExtractive QA
Read More

What I built

  • End‑to‑end ingest → chunk → embed → retrieve → answer pipeline with citations.
  • No paid APIs; runs on a laptop (MiniLM encoder + RoBERTa SQuAD2).
  • CLI with ticker input, doc caching, and top‑k controls.

Links

View Code
Drift-aware fraud detection system

Fraud Detection — Drift‑Aware

Recall‑optimized fraud classifier with prequential (time‑ordered) eval and drift tracking (PSI). Compares static vs adaptive update policies.

Imbalanced LearningRecall@FPRPSI DriftScikit‑learn
Read More

What I built

  • Baselines (LR, RF) + threshold tuning to maximize Recall@FPR for review teams.
  • Prequential splits to simulate streaming; drift quantified via PSI.
  • Optional updating (periodic / drift‑triggered / sliding window) for stability.

Links

View Code
Speech Emotion Recognition MFCC + CNN

Speech Emotion Recognition — MFCC + CNN

Audio ⇢ MFCC feature maps ⇢ CNN classifier with real‑time prediction via mic or .wav upload.

LibrosaMFCCPyTorchGradio
Read More

What I built

  • Signal preprocessing & augmentation; class‑weighted training loop.
  • Confusion‑matrix‑driven error analysis to separate similar emotions.
  • Light Gradio UI for quick demos.

Links

View Code

Resume

Click below to view or download my resume, or browse the summary below.

View Resume (PDF)

⚡ Skills

Programming & ML
PythonRC++ NumPyPandasscikit‑learn PyTorchTensorFlowXGBoostLightGBM
LLMs & NLP
TransformersHugging FaceLangChain RAGFAISSPrompt Engineering Text SummarizationSentiment Analysis
MLOps & Cloud
AWS (S3, Lambda, API Gateway, CloudWatch)SageMaker DockerGitHub ActionsTerraform (IaC) CI/CDMonitoring & Logging
Data & Analytics
EDAFeature Engineering Anomaly DetectionTime‑series Forecasting Drift (PSI)Calibration AB TestingSQL
Web & Visualization
FastAPIFlaskGradioStreamlit PlotlyMatplotlibPower BITableau

💼 Experience

  • Instructional Design Assistant — EdPlus @ ASU Aug 2024 – Present · Scottsdale, AZ
    • Analyzed 10k+ learner interactions to surface insights that improved course outcomes.
    • Built datasets and pipelines for ML experiments; partnered with cross-functional teams.
    • Supported A/B testing and model evaluation to guide course design decisions.
  • Data Analyst Intern — Suzlon Energy May 2021 – Aug 2021 · Pune, India
    • Developed predictive models for wind turbine RUL; improved detection accuracy to ~95%.
    • Preprocessed high-volume SCADA data; optimized features for time-series modeling.
    • Built Tableau dashboards to monitor anomalies and KPIs across turbines.

🎓 Education

  • M.S. in Data Science — Arizona State University · Tempe, AZ
  • PG Diploma in AI/ML — MIT World Peace University · Pune, India
  • B.C.A. in Computer Applications — MIT World Peace University · Pune, India

🚀 Projects

  • NLP App: Sentiment/Emotion/Text Generation using GPT-2 & BERT; Docker + FastAPI.
  • Speech Emotion Recognition (SER): MFCC + spectrogram features → hybrid CNN-RNN; error analysis with confusion matrix for robust emotion classification.
  • Stock.io (Realtime Prediction):Streamlit app: finance news sentiment (web-scraped, 3-day window) + LSTM on OHLC for dual stock recommendations.
  • Predictive Maintenance: Time-series models for turbines; anomaly detection & alerts (Suzlon)

🏅 Certifications

  • HarvardX — High-Dimensional Data Analysis
  • IBM — Data Science / NLP
  • AWS — Machine Learning Specialty (ML‑SC01) Verify
  • AWS — Cloud Practitioner (coursework)
  • Infosys — Applied NLP / Cloud

About Me

My path into data science, from hands-on projects to graduate research and machine learning at scale.

  1. BCA

    🎓 B.C.A. — Bachelor of Computer Applications

    I began with a Bachelor of Computer Applications (programming, DSA, databases). My capstones and an internship at Suzlon sparked my interest in applied data.

    • Internship @ Suzlon: predictive modeling for turbine health, time-series preprocessing, dashboards.
    • GPA: 3.8 / 4.0
  2. PGD

    🧠 PG Diploma in AI/ML

    As AI accelerated, I completed a PG Diploma and assisted on an NLP research capstone in Speech Emotion Recognition (SER), getting hands-on with transformers and modern NLP.

    • Capstone: SER using deep learning & NLP.
    • GPA: 3.7 / 4.0
  3. MS

    📚 M.S. in Data Science (ASU)

    I’m currently pursuing my M.S. at Arizona State University (GPA: 3.6), aiming for ~3.8 by December. Work spans ML systems, statistical modeling, big-data analytics, and cloud.

    • Now: building ML pipelines & model evaluations; supporting A/B testing and analytics.

📘 Relevant Coursework — ASU (taken / planned)

Data Mining (CSE 572) Statistics for Data Analytics (DSE 501) Analyzing Big Data (IFT 511) Design of Experiments (IEE 572) Advanced DB Management (IFT 530) Data Visualization (CSE 578) Prob. & Random Processes (EEE 554) Probability & Stats for Eng (IEE 380)

Focused on ML, stats, data engineering, and scalable analytics. (Fall plan included.)

🧠 Relevant Coursework — PGD AI/ML & B.C.A.

Machine Learning Deep Learning Natural Language Processing Data Structures & Algorithms Database Systems Software Engineering

🌄 Outside Work

  • Swimming & trekking
  • Photography
  • Exploring culinary spots
  • Making explainer videos on new tech

📫 Say Hello

I hope you enjoyed my portfolio. Feel free to reach out, always happy to talk.

Email Call LinkedIn