KU

Scroll to explore

About Me

I'm an AI Engineer and builder with a B.Tech in Engineering Physics from IIT Roorkee (CGPA: 8.415/10). I specialize in turning cutting-edge AI research into production-grade products — from knowledge ingestion pipelines and agentic workflows to multilingual RAG architectures that serve real users.

My work spans the full LLM stack: fine-tuning large models (Llama-3-70B, 90B-Vision), deploying serving infrastructure with vLLM, designing evaluation frameworks, and shipping end-to-end AI systems for healthcare and conversational AI. I think like a founder — every feature I build starts with a user problem, not a paper.

Beyond engineering, I'm drawn to building businesses. I've won multiple startup competitions and business case studies at IIT BHU, Techfest IIT Bombay, and Enactus — spanning social entrepreneurship, electric mobility, and sustainable product design. Currently an AI Engineer at BorderPlus, previously at Convai Technologies.

0 Years of Experience
0 Awards & Wins
0 Projects Shipped

Experience

01

AI Engineer

BorderPlus, Bangalore Nov 2025 — Present
  • Architected a centralized knowledge ingestion and semantic indexing platform for large healthcare corpora with document-specific preprocessing, chunking strategies, and embedding pipelines.
  • Developed multi-step agentic pipelines orchestrating LLM reasoning for document classification, rule abstraction, and spoken clinical transcript processing.
  • Built multilingual RAG architectures combining dense and sparse retrieval, query reformulation, intent classification, and user-memory integration with inline citations.
  • Designed automated compliance evaluation pipelines and improved production conversational agents through prompt orchestration and guardrails (15% token reduction).
02

Data Scientist — Applied Research Team

Convai Technologies Inc., Bangalore June 2024 — Oct 2024
  • Deployed and served LLMs using vLLM on custom cloud infrastructure; built unified backends supporting OpenAI, Anthropic, Gemini, and LLaMA-family models.
  • Implemented hybrid semantic similarity + BM25 + LLM chunking and retrieval, reducing end-to-end latency by 27% and memory usage by 3-4x.
  • Fine-tuned Llama-3-70B and Llama-3.2-90B-Vision models for controllability, factuality, and response consistency.
  • Built research prototypes including multi-agent conversational systems, browser-based autonomous agents via gRPC, and real-time WebRTC plugins.
03

Machine Learning Intern

Convai Technologies Inc., Bangalore Feb 2024 — Apr 2024
  • Designed a probabilistic framework for dynamic personality modeling in LLMs using Big Five personality traits.
  • Built synthetic data generation pipelines to fine-tune smaller LLMs for emotion and context modeling.
  • Developed evaluation metrics and testing frameworks for conversational agent performance.
04

MITACS Globalink Research Intern

McMaster University, Hamilton, Canada May 2023 — July 2023
  • Developed multi-objective optimization models for EV recycling center placement, minimizing financial cost and environmental impact.
  • Conducted comprehensive synthesis of recycling supply chain literature for sustainable infrastructure planning.

Projects

What I've Built

End-to-end projects — from idea to deployed product, with code on GitHub.

Research & Explorations

Academic and research projects that shaped my thinking across AI, physics, and optimization.

Deep Learning

Battery SoC Prediction for EVs

Novel deep learning architecture achieving 99.6% accuracy for State of Charge prediction. Created an India-specific battery dataset and presented at the IMESD Conference.

PythonPyTorchSignal Processing
Deep Learning

Siamese Networks for Signature Detection

Siamese CNN with contrastive loss achieving 81% accuracy on the ICDAR 2011 dataset for signature verification and forgery detection.

TensorFlowCNNContrastive Loss
Physics Simulation

Monte Carlo Light Transport

Physics-based Monte Carlo simulation for photon transport in human epidermal tissue, with strong correlation to published experimental data.

PythonMonte CarloBiophysics
Optimization

EV Recycling Center Placement

Multi-objective optimization models for electric vehicle recycling center placement, jointly minimizing financial cost and environmental impact using non-linear techniques.

PythonOptimizationSustainability
Machine Learning

Student Expenditure Prediction

Designed survey methodology and feature engineering pipelines to predict student financial behavior using SVM with multiple kernel configurations.

PythonSVMFeature Engineering
Quantum Computing

Quantum Music Composer

Winner of Qiskit Fall Fest 2021, IIT Roorkee Quantum Hackathon. Designed a quantum-based music composition system using Qiskit.

QiskitQuantum ComputingPython
Robotics

Swach Bharat Bot

Autonomous cleaning robot built with Arduino. Uses dual IR sensors for obstacle avoidance, a servo-driven sweeping arm, and an LCD for real-time status display.

ArduinoIR SensorsServo MotorC

Skills

Programming & ML

Python SQL PyTorch NumPy Pandas TensorFlow

LLMs & Representation Learning

HuggingFace vLLM LangChain OpenAI Anthropic Gemini ChromaDB

Systems & Infrastructure

Docker GitHub FastAPI Flask PostgreSQL WebRTC Google Cloud

Research & Methods

RAG Architectures Agentic Systems Fine-Tuning Evaluation Frameworks Optimization

Honors & Awards

MITACS Globalink Research Fellow

2023 — Research internship at McMaster University, Canada — multi-objective optimization for EV recycling infrastructure

🏆

IIT Roorkee Encore Award

2023 — Recognized for all-round excellence among a batch of 1200 students at IIT Roorkee

🏆

Winner — Defining Innovation Award

2023 — Social Budgeting Convention, Enactus TIET — Ideated sustainable flower recycling businesses (incense sticks, perfumes, rural tourism) for a village in Bengal

🏆

Runner Up — Start Bucks: Startup Pitching Competition

Aug 2022 — E Cell IIT BHU — Pitched a restaurant analytics and advertisement platform to promote local small businesses and minimize food wastage

🏅

Finalist — Techfest IIT Bombay: Electrifying Transportation

May 2022 — Business development competition in electric mobility — Led team to finals (Top 15 out of 100+ teams)

🏆

Winner — Enact-a-thon: Social Entrepreneurship Case Study

Feb 2022 — Enactus IIT Roorkee — Designed a self-sustainable education platform providing free mentorship and guidance for JEE-NEET aspirants

🏆

Winner — Qiskit Fall Fest: Quantum Computing Hackathon

Nov 2021 — Quantum Computing Group IIT Roorkee — Built a quantum-based music composer using Qiskit over a 3-day hackathon

Get in Touch

I'm always open to discussing new opportunities, startup ideas, research collaborations, or interesting problems at the intersection of AI and building products.