training portfolio.model epoch 000/100
loss epochs
loss 2.4031 click to skip →
learning complete
accuracy0.000 PR-AUC0.000 F10.000 val_loss0.0000
~/anubhav.tiwari

Anubhav Tiwari

Senior Machine Learning Engineer · five years shipping production ML

Resume
// 01  about

I build production ML
systems that actually ship.

Most ML work dies between the notebook and prod. Mine doesn't. Five years of owning the full cycle — from data pipeline design through model development, cloud deployment, and monitoring in production. I work across computer vision, audio signal processing, NLP, and MLOps, and I treat the engineering quality of the system as seriously as the model accuracy.

Computer Vision

Crowd density estimation, object detection, license-plate recognition, face recognition. Real-time video pipelines at production scale.

Audio & Signal

Voice activity detection, speaker diarization, noise-robust systems. Signal-level engineering in real-world noisy environments.

NLP & Code Gen

Natural-language-to-code models, semantic understanding, research prototypes. Published papers in deep learning for code.

defpredict(x):
// 02  experience.log

A five-year trajectory shipping ML at scale

From research labs to production cloud — four roles, one through-line: own the lifecycle.

2023 — Present

Quantiphi Analytics

Senior Machine Learning Engineer
CV Audio Cloud
 hover for details

Shipping production CV & audio systems end-to-end — from crowd density to license plates to speaker diarization.

  • Crowd density estimation models holding 95%+ accuracy in varied lighting
  • License plate recognition pipelines on real-time video streams
  • Voice activity detection hardened for noisy production audio
  • Speaker diarization models deployed across cloud infrastructure
CV Audio Cloud
2022 — 2023

Quantiphi Analytics

Machine Learning Engineer · L2
CV Face Rec AWS / GCP
 hover for details

Leading ML initiatives from POC through production deployment for enterprise clients.

  • Established face recognition capabilities for enterprise clients
  • Built object detection models with high precision in real-world scenes
  • Deployed ML solutions across AWS and GCP infrastructure
  • Owned the lifecycle from data through monitoring
CV Face Rec AWS / GCP
2021 — 2022

Verifyer Technologies

Machine Learning Engineer
Analytics Modeling
 hover for details

Designed ML models for varied business problems — data exploration through operational deployment.

  • Designed models tailored to specific business problems
  • Analyzed large datasets to surface actionable insights
  • Built high-accuracy prediction models for operational use
  • Delivered data-driven solutions with cross-functional teams
Analytics Modeling
2019 — 2020

RMgX Technologies

Machine Learning Engineer
NLP Research
 hover for details

Started in NLP research — published papers, prototyped novel architectures, mentored juniors.

  • Conducted research in natural language processing
  • Published papers on innovative deep learning models
  • Implemented prototype solutions for academic research
  • Mentored junior researchers in ML methodologies
NLP Research
// 03  selected_work

Six shipped systems, each with a live demo

From crowd-density estimation to live MLOps pipelines — open any card to play with the real thing.

P.01 · computer vision · pytorch

A real-time crowd density model that holds 95%+ accuracy across varied lighting conditions.

Density mapping for public spaces — built end-to-end from data pipeline through deployment.

PyTorch CV Real-time
P.02 · audio · signal processing

Speaker diarization that figures out who-spoke-when in noisy real-world audio.

Separating overlapping voices into clean speaker timelines.

Audio Signal
P.03 · computer vision · ocr

License plate recognition for live video streams — detect, segment, read.

Bounding-box detection plus OCR on a single pipeline.

CV OCR
KA·03·MJ
P.04 · nlp · code generation

Natural-language-to-code: describe what you want, get executable Python back.

A research-grade NLP system that grew out of published papers in deep learning for code understanding.

NLP TensorFlow Research 
# tiny llm autocomplete
def predict(prompt, model):
    tokens = tokenize(prompt)
    return model.generate(tokens)
P.05 · audio · dsp

Voice activity detection that segments speech from silence in real time.

A lightweight model that runs on edge devices with minimal latency.

Audio DSP
P.06 · mlops · cloud

CI/CD for ML on AWS — notebook to production with monitoring built in.

End-to-end pipelines from data ingest through canary deploy with rollback.

Docker AWS CI/CD
// 04  stack

The toolkit behind the work

model_card.yaml v5 · prod
domains computer vision audio & signal nlp mlops
languages Python C++ JavaScript SQL Bash
frameworks PyTorch TensorFlow Keras OpenCV scikit-learn NumPy Pandas Librosa
infra AWS GCP Docker Kubernetes Azure CI/CD
tooling Git MLflow CUDA Jupyter Linux MySQL
status 5+ years · production scale · open to opportunities
// 05  connect

Got a hard ML problem? Let's talk.

~/anubhav — zsh
email
abt.exp@gmail.com
github
github.com/abtExp
linkedin
linkedin.com/in/abtExp
x · twitter
@overfitowl