Skip to content

Jersey City, NJ · Open to AI engineering roles

Mayank Kumar Pokhriyal

AI Engineer · Agentic AI · RAG · LLM Systems

AI Engineer and Data Scientist with 5+ years of professional experience at IBM and Kyndryl, and a current M.S. student in Artificial Intelligence graduating December 2026. I specialise in agentic AI systems, retrieval-augmented generation, LLM-powered APIs, and end-to-end AI product development.

Download résumé ↓See the work →
Mayank Kumar Pokhriyal
Years in production ML
5+Years in production MLIBM → Kyndryl, 2019–2025
Verified certifications
34Verified certificationsAnthropic, IBM, Kyndryl, Interskill
WESAD test accuracy
94%WESAD test accuracyCNN–BiGRU–Attention, multimodal
M.S. AI graduation
Dec 2026M.S. AI graduationYeshiva University, Katz School

01The path

From data pipelines at IBM to agentic AI systems.

Five years building ML that had to stay up in production, and a master's to go deeper on the theory underneath it.

Professional experience

  1. Jun 2026 — Present

    Dispatch Energy

    Full-Stack AI Engineer (Internship)

    • Designed and built the authentication and security architecture for an enterprise MCP (Model Context Protocol) gateway that connects Claude and other AI clients to Quickbase — OAuth 2.0 authorization flows, Microsoft Entra ID token validation, per-user token storage backed by Azure Key Vault, and JWT-based MCP session auth with self-service onboarding.
    • Found and fixed real security holes in that gateway before it reached production: a Key Vault secret-name collision that could leak one user’s tokens to another, and an over-broad tool surface that exposed schema, admin and delete operations to any connected model. Added structured audit events so every tool call is now accountable.
    • Built an MCP-based RPA system driving BESS (battery energy storage) interconnection studies on PowerVAL — Playwright workflow discovery, guarded submission flows with validation, release-policy tool gating, and a supervised Windows worker runtime on Azure.
    • Engineered a cycle-aware battery degradation and revenue-modelling engine for the solar-plus-storage pricing platform, and split the product into dedicated asset-revenue and customer-savings experiences. Fixed utility billing proration bugs that were skewing savings figures.
    • Shipped the delivery pipeline behind it all: GitHub Actions → Azure Container Registry → Azure Container Apps, with canary traffic-splitting and smoke-test gating on staging before any production promotion.

  2. Feb 2026 — Jun 2026

    Katz School of Science & Health

    Yeshiva University

    Research Assistant · Manhattan, NY

    • Designed the two-agent architecture behind Anida, a clinical-trial matching agent that reasons across 40,000+ trials — splitting a single voice assistant into a ConversationOrchestrator and a BackendSpecialist, shipped behind a feature flag so single-agent mode kept working throughout.
    • Built the patient contact and consent flow, and added trial-information guardrails to the agent prompts so the system cannot assert anything about a trial that is not in the retrieved record — the safety constraint that matters most in a healthcare setting.
    • Replaced the realtime voice stack with a Whisper speech-to-text and REST text-to-speech pipeline over Twilio phone streams, adding voice-activity detection, session management and SMS handoff, plus a browser voice widget.
    • Extended clinical trial search to accept NCT identifiers, added prompt-size validation, and wrote the developer guide along with unit tests covering agent handoff and session lifecycle.

  3. Sep 2021 — Jan 2025

    Kyndryl

    AI/ML Engineer · Bangalore, India

    • Designed, trained and deployed large-scale fraud detection and credit-risk prediction systems using TensorFlow and XGBoost for high-volume financial transaction data.
    • Built and maintained production-grade RESTful ML inference services in Python and FastAPI, with modular architecture and authentication.
    • Automated data ingestion and transformation pipelines processing millions of records per day, significantly reducing operational latency.
    • Worked with cloud, DevOps and data engineering teams to containerize and deploy ML systems using Docker and CI/CD pipelines.
    • Delivered explainable AI outputs and monitoring dashboards for business-critical risk systems.

  4. Aug 2019 — Sep 2021

    IBM

    Data Engineer · Bangalore, India

    • Developed and optimized large-scale ETL pipelines for enterprise financial datasets using Python, SQL and workflow schedulers.
    • Built analytical reporting dashboards giving insight into multi-million-dollar operational and compliance systems.
    • Partnered with ML teams to productionize predictive models into business data platforms.

Education

  1. Jan 2025 — Dec 2026

    Yeshiva University

    Katz School of Science & Health

    M.S. Applied Artificial Intelligence

    Manhattan, NYGPA 3.7 / 4.0

  2. Sep 2020 — Sep 2022

    BITS Pilani

    Birla Institute of Technology and Science

    M.Tech Data Science & Engineering

    IndiaGPA 7.5 / 10.0

  3. Aug 2015 — Aug 2019

    Dr. A.P.J. Abdul Kalam Technical University

    B.Tech Computer Science and Engineering

    IndiaGPA 7.2 / 10.0

02What I do

I build AI systems that survive contact with real traffic.

Right now I build MCP infrastructure connecting AI agents to enterprise systems at Dispatch Energy — auth, security and the delivery pipeline around it. Before that, healthcare-adjacent predictive modelling, multimodal deep learning and cloud-deployed ML at IBM and Kyndryl. I care about systems that survive contact with real traffic, not notebooks that only run once.

Platform experience

  • Building end-to-end LLM systems using SDKs, CLI tools and local inference servers (Ollama).
  • Vector databases (FAISS, Chroma), embeddings and semantic search pipelines for RAG applications.
  • RESTful AI microservices with authentication, persistent storage and cloud-ready modular architecture (FastAPI + Docker + AWS).

Stack

AI & LLMs
Agentic AILangChainOllamaOpenAI APIRAGPrompt EngineeringEmbeddingsVector Databases
Machine & Deep Learning
PyTorchTensorFlowCNNsRNNsAttentionTransformersMultimodal Learning
Backend
PythonFastAPIFlaskREST APIsJWT AuthSQLAlchemy ORM
Frontend
React.jsComponent-based UIAPI integration
Databases
PostgreSQLSQLiteFAISSChromaDB
Cloud & DevOps
AWS EC2AWS S3DockerGitHub ActionsNginx

03Selected work

Six projects worth your time.

Local-first agents, retrieval pipelines over documents nobody wants to read, and deep learning on wearable sensor data.

Agentic & RAG

on résumé

Agentic AI Profile — open on GitHub

A local agent that answers recruiter questions about me, and logs the lead.

A fully local Streamlit agent powered by Ollama that lets a recruiter interview my professional profile conversationally. Reasoning runs through an orchestrator that emits deterministic JSON tool calls against a structured profile store, which is what keeps it from hallucinating my background. When it detects genuine hiring intent it captures the lead to Google Sheets automatically.

  • Python
  • Ollama
  • llama3
  • Streamlit
  • Tool calling
  • Google Sheets API
GitHub ↗PythonUpdated Dec 2025

Agentic & RAG

Domain-Specific RAG Agent — open on GitHub

Production RAG with agent-controlled retrieval and an explicit refusal path.

A local-first RAG service: upload a PDF, and it chunks with metadata preserved, embeds via Sentence-Transformers, and persists a FAISS index to disk. The part I care about most is the agent layer, which decides whether to retrieve or to refuse — grounding answers in retrieved context rather than guessing when the corpus has nothing relevant. Fully offline through Ollama, exposed over FastAPI with an optional Streamlit UI.

  • FastAPI
  • FAISS
  • Sentence-Transformers
  • Ollama
  • Streamlit
  • Python 3.13
GitHub ↗PythonUpdated Dec 2025

Agentic & RAG

on résumé

RAG Financial 10-K Analyzer — open on GitHub

Natural-language querying over SEC 10-K filings, entirely on local inference.

An end-to-end retrieval-augmented generation pipeline built for documents far longer than any context window: SEC 10-K filings. Handles document chunking, FAISS vector indexing and semantic retrieval, then answers with local LLM inference so nothing about the filings leaves the machine.

  • FAISS
  • Ollama
  • LangChain
  • Python
  • Jupyter
GitHub ↗Jupyter NotebookUpdated Nov 2025

MLOps & Systems

on résumé

LLM-Powered Sentiment API — open on GitHub

Containerized FastAPI sentiment service with CI/CD and structured exports.

A sentiment analysis service built the way a production service should be: LangChain and Ollama behind REST endpoints, structured outputs rather than free text, and automated CSV/PDF/DOCX export. Shipped with Docker and a GitHub Actions CI/CD pipeline.

  • FastAPI
  • LangChain
  • Ollama
  • Docker
  • GitHub Actions
GitHub ↗PythonUpdated Nov 2025

Healthcare AI

on résumé

Multimodal Stress Detection — open on GitHub

CNN–BiGRU–Attention over wearable ECG, EDA and EMG signals.

A healthcare-focused deep learning model classifying physiological stress from wearable sensor data in the WESAD dataset. Built as research rather than a demo: the repo includes ablation studies, confusion-matrix analysis and research-style reporting alongside the model.

94%test accuracy

  • PyTorch
  • CNN
  • BiGRU
  • Attention
  • Multimodal
  • Jupyter
GitHub ↗Jupyter NotebookUpdated Dec 2025

Healthcare AI

on résumé

Circadian Rhythm Optimizer — open on GitHub

Predicts your body clock phase, then plans a 7-day realignment to solar time.

A modular deep-learning pipeline over the MMASH dataset that ingests sleep diaries, actigraphy, RR-interval and questionnaire data to estimate circadian phase. A CNN→BiGRU predicts phase encoded as (sin φ, cos φ); pvlib supplies true sunrise, solar noon and solar midnight for any coordinate, and a recommendation engine turns the phase difference into an actionable week-long light-and-sleep plan.

  • PyTorch
  • CNN
  • BiGRU
  • pvlib
  • MMASH
  • Python
GitHub ↗PythonUpdated Nov 2025
All 12 curated projects

04Credentials

34 certifications, every one independently verifiable.

Issued by IBM, Kyndryl, Interskill, Cognitive Class and Udemy between 2019 and 2024. Each one links to its official record — please check them.

22

AI & Data Science

4

Cloud & DevOps

8

Enterprise Systems

Browse and verify all 34