Open to AI & Forward-Deployed Engineer roles · Mumbai

Engineering Intelligence.
Designing Mobility.

Building agentic AI for consumer finance — voice agents and orchestration for collections calls. I specialise in MCP-orchestrated pipelines, production RAG, and LLM systems that reason, retrieve, and act autonomously — with the same precision-engineering mindset I bring to automotive architecture.

kbedi03@gmail.com
+91 7021484750
Mumbai, Maharashtra

What I Build

AI / ML Projects

All 11 projects
Agentic AI

Agentic RAG Pipeline

LangChain Agents, ChromaDB & Llama Fine-Tuning

Fine-tuned Llama on 500+ examples · 100% local GPU inference · zero cloud API cost · MLflow-tracked

LangChain AgentsChromaDBRAGLlama (Fine-Tuned)CUDA+5
Automotive AI

Geospatial Predictive Maintenance

Anomaly Detection on Commercial Fleet Telemetry

F1 0.82 on the minority fault class · SMOTE-balanced · interactive geospatial risk maps

PythonScikit-learnSMOTEPandasNumPy+5
Computer Vision

Breast Cancer Detection

Medical Imaging CNN Pipeline

96.5% validation accuracy · CUDA-accelerated training · Streamlit inference UI

PythonTensorFlowKerasCNNCUDA+4
GenAI

AI Color Palette Generator

Computer Vision Pipeline + Gemini Vision API

Image → branded hex palette in one pipeline · shipped in <2 weeks · Dockerised REST API

PythonFastAPIOpenCVGemini Vision APIDocker+2
Reinforcement LearningPublished Research

Field Trip Fighters — Adaptive RL Boss AI

Hybrid Q-Learning + PPO · Published Research Paper

95% win vs beginner AI, 72% vs advanced · 2,590 training episodes · peer-reviewed paper

PythonGodot EnginePPOQ-LearningN-gram Pattern Recognition+3
Computer Vision

AI-Based Timetable Detection

Computer Vision + Gemini Vision API

Unstructured image → structured schema · End-to-end automation · Reliable post-processing validation

PythonGemini Vision APIComputer VisionImage PreprocessingPrompt Engineering+2
ML · FinTech

Loan Default Prediction

ML Risk Scoring Model — Inditrade Capital

78% precision on high-risk accounts · +30% risk identification · shipped to the collections team

PythonRandom ForestsScikit-learnSMOTEPandas+5

Experience

Agent Engineer

Prodigal

May 2026 - Sep 2026

• Built production voice AI agents for US consumer-finance clients — multi-LLM orchestration on Vertex AI (Claude Sonnet for reasoning, Claude Haiku for deterministic RPC flows), with Twilio, TCN, and LiveKit telephony, Deepgram STT, and ElevenLabs TTS.

• Cut agent response latency >33% (3s → sub-2s) and worst-case latency 50% (6s → 3s).

• Built an evaluation suite on Cekura MCP to simulate calls and score agent behavior, tracing quality and failure modes across Langfuse, CloudWatch, and TrueFoundry.

• Supported live C-suite client demos alongside the GTM team, contributing to two signed enterprise deals.

Senior Assistant Engineer, AI & ML

Samsung India, Noida

Jan 2026 - Apr 2026

• Built and deployed 3 LLM-powered workflows from requirements through production, cutting manual reporting effort >60%.

• Designed a RAG pipeline integrating internal knowledge bases with LLM inference, reducing hallucination rate ~35% across 50+ evaluation cases.

• Built automated ETL pipelines processing 10,000+ daily records, exposed via FastAPI REST APIs and shipped with CI/CD through GitHub Actions.

• Nominated for Samsung's advanced AI training program in Vietnam — a selective initiative for the top 2% of AI engineering interns nationally.

Technical Arsenal

LLM Orchestration (multi-model)Agentic AI SystemsMCP (Model Context Protocol)Production RAGLLM Evaluation (Langfuse, Cekura)Voice AI (LiveKit, Twilio, TCN)Speech (Deepgram STT, ElevenLabs TTS)Vertex AI · Claude (Sonnet / Haiku)Prompt & Context EngineeringPython for AI/MLFastAPI & REST APIsETL & Data PipelinesObservability (CloudWatch, TrueFoundry)CI/CD (GitHub Actions)Machine Learning (Supervised & Unsupervised)Git & Docker

My Leadership Journey

President

4C Marketing Cell, NMIMS MPSTME

2025 - Present

Lead a 100-member club delivering consulting-style marketing events where companies present real business challenges and student teams develop and present actionable solutions.

Vice President

Innovation and Entrepreneurship Cell, NMIMS MPSTME

2024 - 2025

Led an 18-member cross-functional team to build and launch iecnmims.com, automate club workflows, and produce the Innovision podcast with partner brands.

Head of Digital Creatives

Taqneeq 17.0

2023 - 2024

Directed digital branding and social media strategy, establishing a cohesive visual identity and managing Instagram content and brand collaborations for the university tech fest.

Head of Technical Department

Innovation and Entrepreneurship Cell

2023 - 2024

Led Next.js development and GitHub-based collaboration, mentoring the technical team across two cohorts.

Education

B.Tech in Artificial Intelligence

CGPA 3.19/4.0

MPSTME, NMIMS University | 2021 - 2026

Prabhavati Padamshi Soni International Junior College (A Levels)81.6% | 2019-2021
Vibgyor High Goregaon West (IGCSE)89.875% | 2019