Building production-ready AI systems that deliver real business impact →
Abhisek Bose

Product Manager | Gen AI | LLMOps | AI Engineer

Building
Production-Ready
AI Systems.

Agentic AI Hybrid RAG Model Routing LLMOps Reliability

I design and ship agentic AI products with robust architecture, hybrid retrieval, tool orchestration, and LLMOps controls for reliability, cost, and scale. PM strategy plus hands-on builder execution from prototype to production rollout.

View Projects Contact Me

Previously At

Codeclouds Tier5 Arodek Kutchina
Abhisek Bose portrait

By the Numbers

14+ years. Measurable AI product impact.

$6M+

GMV in Year 1

Scaled an AI-augmented checkout SaaS to 15 enterprise merchants.

$2.4M+

ARR Influence

Drove SaaS + transaction pricing outcomes with 110% NRR.

50+

Experiments Run

A/B and analytics loops across checkout and growth workflows.

40%

Conversion Lift

Checkout conversion improvement through rapid product iteration.

30%

Chargebacks Reduced

AI-enhanced fraud protection with safety and reliability guardrails.

12+

AI Systems Built

Production systems across agentic AI, hybrid RAG, routing, and model training pipelines.

Experience

Where I've shipped.

Current Sep 2024 — Present

Codeclouds IT Solutions

Product Manager

India / Global Teams

0-to-1 AI-augmented commerce and operations products with production-grade reliability and measurable business outcomes.

  • Led AI-augmented custom checkout SaaS to product-market fit with 15 enterprise merchants and $6M+ GMV in year one.
  • Contributed to $2.4M+ ARR and 110% net revenue retention through SaaS + transaction pricing strategy.
  • Drove 50+ experiments with 40% conversion lift, 25% AOV gain, and 35% cart-abandonment reduction.
  • Integrated AI fraud protection to reduce chargebacks by 30% while maintaining 99.9% uptime standards.
  • Shipped LLM-powered recruitment workflows with PII masking, 35% parsing accuracy lift, and 40% faster onboarding.
Growth Jul 2019 — Aug 2024

Tier5 Technologies

Senior Product Manager

India

Built CRM and organic marketing products with AI-enabled support workflows and outcome-led delivery.

  • Identified and shaped a $5B+ TAM opportunity across CRM and organic marketing product lines.
  • Defined product vision, GTM strategy, and OKRs to align engineering, UX, and commercial teams.
  • Designed and shipped a RAG-based customer support assistant improving CSAT by 15%.
  • Reduced churn by 10% while owning end-to-end lifecycle for products including Friender and Chirply.
Enterprise Aug 2017 — Jun 2019

Arodek Technologies

Principal Consultant Lead

India

Delivered large-scale B2B CRM capabilities by combining product strategy, consulting, and technical execution.

  • Drove $15M+ revenue in the first year from two major CRM feature launches.
  • Led cross-functional delivery across 15 engineers, 3 UX designers, and sales stakeholders.
  • Improved product quality by 15% via OKR-led prioritization and roadmap governance.
  • Improved project profitability by 60% through technical pre-sales consulting and scope clarity.

Portfolio

Case studies that show the work.

Compliance Brain

›

CortexBrain

Auditable, self-correcting enterprise AI knowledge system built for compliance-heavy environments.

Audit-ready AIConfidence gatingVersioned corrections

AgRAG

›

Agentic RAG Chatbot

Multimodal enterprise RAG system with orchestrator, hybrid retrieval, tool routing, and feedback loop.

Vector + graph + imageTool routingReflection loop

CCIL

›

ChampIQ / CheckoutChamp Agent

Multi-agent conversational analytics and operations layer over CheckoutChamp APIs using Gemini, MCP, and LangGraph.

70+ MCP API toolsApproval gatesSSE streaming

LLMOps

›

LLM AI Router

Dynamic model-routing engine selecting providers by cost, speed, and quality with guardrails and observability.

Cost-aware routingCaching + fallbacksProvider orchestration

AI Product Compass

›

AI Product Copilot

AI platform for product teams covering PRDs, prioritization, GTM planning, and PM coaching workflows.

PRD generationPrioritization workflowsGTM planning

WhyWeDecided

›

Meeting Journal

Local-first desktop meeting intelligence app for recording, summary, action extraction, and decision memory.

Transcription + summariesAction extractionSearchable memory

Issue0

›

Issue0 Studio

Agentic QA platform for intent-based, self-healing automation and resilient Playwright test generation.

Intent-driven testingPlaywright automationMCP workflows

DesignMaster

›

Visual AI Re-Imaginer

Structured website reconstruction and AI editing workflow with safe normalization and clean export-ready code.

Safe normalizationLive iframe editingZIP export

Llama RAG

›

Llama RAG Advanced

Advanced PDF RAG prototype using local Llama inference with retrieval quality optimization pipeline.

4-bit local inferenceReranking pipelineGrounded Q&A

Model Training

›

SLM Self-Supervised

From-scratch next-token language model training experiment using custom PyTorch loops.

Causal maskingTransformer internalsTraining loops

Model Adaptation

›

SLM Instruction FT

Instruction-style fine-tuning workflow for compact transformer models on structured Q&A data.

Instruction tuningTarget maskingInference checks

Side Projects

Things I've shipped.

Products built end-to-end using AI tools, all tied to real user workflows.

11+

featured AI builds

Issue0 Studio

Agentic QA platform for intent-based, self-healing automation and resilient Playwright test generation workflows.

Next.js 15PlaywrightMCP

3

model-training notebooks

SLM + Advanced RAG

Self-supervised training, instruction fine-tuning, and retrieval optimization over local Llama inference.

PyTorchTransformersLlama 3.2 (4-bit)

1

desktop intelligence app

Meeting Journal

Local-first meeting intelligence product for recording, transcription, summary, action extraction, and searchable decision memory.

ElectronSQLiteGemini

Skills

Product + AI execution stack.

Product Management

0-to-1 Product DevelopmentRoadmap OwnershipPRD & Problem FramingGTM StrategyPricingOKR / KPI Design

Discovery & PM Research

Customer InterviewsJourney MappingUsability TestingExperiment DesignFunnel AnalysisCross-functional Alignment

GenAI & LLM Engineering

OpenAI GPT-4/4oGeminiClaudeLlama 3.2Tool CallingMultimodal AI

LLMOps & Reliability

Dynamic RoutingCachingRetries + FallbacksConfidence ScoringHallucination MitigationLangfuse

RAG & Knowledge Systems

Hybrid RetrievalVector + Graph + MetadataQuery RewritingRerankingSemantic MemoryAudit Trails

Model Training

PyTorchHugging FaceSelf-Supervised TrainingInstruction FTTransformer Internals4-bit Quantization

Full-Stack AI Delivery

Python + FastAPINode.js + Bun + HonoNext.js + React + TypeScriptMongoDB + Neo4j + PostgreSQLRedis + SQLite + Qdrant/LanceDBDocker + Celery + S3

Education

Postgraduate Studies in Computer Applications

Punjab Technical University

As per CV

Undergraduate Studies in Computer Applications

Punjab Technical University

Certifications: Bright Idea Award (2024), Enterprise Agile, Product & Python tracks

As per CV

Beyond Product

From AI experiments to production systems.

I help teams move from demos to reliable AI products with lower operational cost, stronger governance, and measurable business outcomes.

50%

Reduction in LLM drift/safety incidents via routing and monitoring controls.

35%

Improvement in resume parsing accuracy for AI-assisted recruitment workflows.

40%

Faster onboarding from LLM-powered automation with guardrails and PII-aware design.

150+

PMs mentored across coaching and portfolio-review programs.

Let's Work Together

Let's build AI that works in production.

If you are hiring for Gen AI, LLMOps, or Applied AI engineering, I would be glad to connect and discuss how I can help your team ship reliably.

Contact Me → LinkedIn ↗
Gen AI Engineering·LLMOps Engineering·Applied AI Engineering·Open to Remote / Global Roles