~/jaiprasad_ status: training

// data scientist · ml engineer · 8+ yrs

Training models
that ship to prod.

I'm Jaiprasad. I build LLM agents, recommendation engines, computer‑vision pipelines and the MLOps platforms that keep them running. From research notebook to production, across healthcare, security and e‑commerce.

Get in touch → resume.pdf ↓

jprampure@gmail.com · +91 99012 75669 · Abu Dhabi / Bengaluru

  • 0years in AI / ML
  • 0EBITDA uplift · pricing
  • 0AWS cost reduction
  • 0WBC detection rate

01About

Seasoned full‑stack data scientist with 8+ years of experience building AI solutions and leading cross‑functional teams through end‑to‑end delivery. I have managed research and data science groups, driven annual planning, product design and ML architecture, and advised stealth and early‑stage AI startups on stack selection and scalable design.

I care about translating hard technical ideas for mixed audiences and aligning AI work with what the business actually needs. Currently based between Bengaluru and Abu Dhabi, working as an individual contributor on the Mensa Brands data science team.

objectivemaximise business impact
policyship, measure, iterate
explorationstealth‑stage AI advisory
reward signalCTR · EBITDA · latency · cost

02Experience

  1. Nov 2022 — Present

    Mensa Brands Data Science Team (IC) · Bengaluru / Abu Dhabi

    • Content to Commerce Platform. Built an in‑house advertising ecosystem that tags products inside articles. Designed a recommendation engine with offline training and online serving on FAISS vector search and OpenAI text embeddings, lifting CTR from 0.3 to 0.47. Added an internal RAG system that drafts new articles from the product and article database.
    • Pricing Central. Dynamic pricing system using Prophet forecasting and Mixed Integer Programming to optimise prices across the portfolio, delivering a 5‑7% EBITDA uplift.
    • Infinity. Serverless MLOps platform on AWS CDK integrating Lambda, Step Functions, Batch and GitHub Actions for automated deployment, training, monitoring and ingestion.

    FAISSRAGProphetMIPAWS CDK

  2. Mar 2021 — Aug 2022

    Johnson Controls Principal Data Scientist · Bengaluru

    • Business Logic Module. Re‑engineered RabbitMQ event processing and derived‑metrics analysis for a 10x faster response time and 15% lower deployment cost.
    • ACVS Engine Revamp. Moved object detection to YOLOv8 with multi‑class, multi‑object tracking on Deep‑SORT, as part of the AI architecture team.
    • JCI AI Roundtable, India. Contributor to the AI Leaders Committee shaping next‑generation applied AI and the long‑term technology roadmap.

    YOLOv8Deep‑SORTRabbitMQAI architecture

  3. Jan 2019 — Mar 2021

    Mfine (Novocura) Senior Data Scientist · Bengaluru

    • Enigma ML Platform. Designed MLOps infrastructure from scratch on Metaflow, MongoDB, Docker, Kubernetes and DVC: one‑click deployment, end‑to‑end tracking, 50% faster retraining and 75% lower AWS spend.
    • Maya AI Assistant. Doctor assistant using Bayesian inference for symptom matching and Siamese networks for differential diagnosis, cutting physician workload 30% with 80% CTR on top‑5 predictions.
    • Medical Embeddings. Trained domain‑specific Word2Vec, GloVe and Bio‑BERT on proprietary data including SNOMED‑CT and ICD‑10, improving every downstream clinical NLP task.
    • Explainable AI library with Grad‑CAM, CAM and attention inference, and mentored interns on MFCC/spectrogram heartbeat classification.

    MetaflowKubernetesBio‑BERTSiamese netsXAI

  4. Dec 2016 — Jan 2019

    SigTuple Data Scientist IV · Bengaluru

    • Shonit blood analyser. Computer vision pipeline for white blood cell extraction (segmentation, adaptive thresholding, morphology) raising sensitivity 33% to a 99% detection rate. Trained 200+ deep models reaching 96% F1 on WBC classification, feeding two clinical studies.
    • Drishti retinal screening. Re‑engineered the diabetic retinopathy pipeline with a feature‑pyramid‑pooling CNN that beat prior benchmarks.
    • Dipstick analyser on ORB descriptors and SVM, plus contributions to the internal training platform for segmentation and structured data.

    Computer visionCNNSegmentationClinical studies

03Notable projects

llm agents · drizz

QA Co‑pilot LLM Agent

Multimodal LangChain agent with custom tools for text interpretation, candidate extraction and a maths tool, driving GPT‑4o to interpret user flows and validate mobile app functionality tests end to end.

LangChain · GPT‑4o · agentic tools

generative video · unscript.ai

Personalised AI Video Generation

Modularised the DeepFaceLab codebase for a 5x faster inference path and integrated Wav2Lip for precise lip‑sync, enabling real‑time personalised video synthesis.

DeepFaceLab · Wav2Lip · PyTorch

computer vision · frshr technologies

Gait Recognition System

Scalable real‑time surveillance system that identifies people at a distance from walking patterns alone, combining OpenGait, YOLOv7 detection and DeepSORT tracking.

OpenGait · YOLOv7 · DeepSORT

04Skills

languages

Python, C/C++, Java, CUDA, MATLAB, Bash, LaTeX, Git, Vim

ml & serving

PyTorch, TensorFlow, OpenCV, HuggingFace, vLLM, Nvidia Triton, FAISS, LangChain, LlamaIndex

mlops

Docker, Kubernetes, Kubeflow, MLflow, Metaflow, DVC, Jenkins, FastAPI, Flask

cloud

AWS (CDK, Lambda, Step Functions, SageMaker, Batch, EC2, ECR, SQS/SNS, Bedrock) and GCP/Azure equivalents

domains

Computer Vision, NLP, Deep Learning, LLMs & LLMOps, Generative AI, Recommendation, Forecasting & Optimisation

05Education

06Patents & publications

07Let's talk

Open to senior data science and ML engineering roles, advisory work with early‑stage AI teams, and interesting problems in general.

jprampure@gmail.com +91 99012 75669 github linkedin