Optum
Nov 2025 - Present
Lead AI/ML
Leading AI engineering for healthcare applications with focus on production reliability, governance, and measurable outcomes.
Profile Brief
AI Architect with 10+ years across IBM Research, Accenture Labs, Samsung Research, True ValueHub, and Optum, delivering research-to-production systems in GenAI and Computer Vision.
Patent-backed innovator (17 filed, 14 granted) focused on scalable enterprise AI, reliable deployment, and measurable business outcomes.
AI Leadership PortfolioProduction AIPatent-led R&D
Jhilam Bera
AI Architect for GenAI, Computer Vision, and Agentic AI Systems
I help teams translate research-grade AI into reliable product systems: architecture reviews, PoCs, model pipelines, evaluation loops, and multi-agent workflows that can move from demo to deployment.
17
Patents filed
14 granted across applied AI and systems
10+
Years in AI R&D
IBM Research, Accenture Labs, Samsung Research
6
Delivery themes
GenAI, CV, Green AI, CAD, robotics, agents
Architecture Focus
LLM platforms, RAG systems, AI evaluation, computer vision deployment, and agentic workflows with governance.
Consulting Outcomes
Clear technical roadmap, implementation risk map, prototype scope, and production readiness checkpoints.
Experience
Research and enterprise delivery journey
True ValueHub
Apr 2024 - Nov 2025
Director AI/ML
Built CAD-to-vision embedding pipelines and LLM-driven cost analysis workflows for manufacturing sourcing decisions.
Samsung Research
Sep 2022 - Jan 2024
Sr. Chief Research Engineer
Built bokeh generation pipelines using segmentation and matting; contributions shipped on Samsung A and M series devices.
Accenture Labs
2020 - Jun 2022
R&D Associate Principal
Led Green AI research to estimate energy/carbon footprint trade-offs for ML workloads across AWS infrastructure.
IBM Research
2014 - 2020
Software Development Engineer / Sr. Inventor
Built applied ML and computer vision systems, including multi-node pill detection and media understanding pipelines.
Projects
High-impact delivery, non-confidential summaries
Case Study
AI Bokeh Effect
Semantic segmentation and matting based bokeh generation pipeline optimized for mobile deployment.
Case Study
CAD Vision Embedding System
Generated multi-angle CAD views and built custom multi-channel vision embeddings for vector database retrieval.
Case Study
Cost Prediction LLM
Used compact transformer workflows for material-cost document parsing, feature extraction, and should-cost prediction.
Case Study
Green AI Research
Modeled utilization-energy-carbon relationships across cloud instance types to improve infrastructure decisions.
Case Study
Chaos Robo Robustness
Simulated robotics perturbations in ROS/Gazebo environments to estimate robustness under adverse operating conditions.
Case Study
Multi-Modal Enterprise Vision
Built production-oriented computer vision components with scalable serving patterns and enterprise integration.
Personal Projects
Community-facing work on social platforms
WhatsApp AI Assistant
AI-first WhatsApp enhancement layer for smart replies, context-aware response drafting, and behavior-driven conversation support.
Auto ReplyBehavior AnalysisOpenAI IntegrationConversation Intelligence
Instagram AI Content Lab
Graph intelligence dashboard for Instagram ecosystems, visualizing user communities, target nodes, mutual relationships, and AI-guided prioritization.
Graph IntelligenceCommunity ClustersAI PrioritizationNeo4j + NetworkX
Patents
Applied invention portfolio
Total Patents
17
Granted
14
| Patent Title |
|---|
| Determining shelter areas for two-wheeler vehicles |
| Finetune image feature extraction using environmental data |
| Personalized timeout control |
| Non-resource-intensive object detection |
| Machine learning based video compression |
| Video compression using cognitive semantics object analysis |
| Object recognition with dynamic switching of visual recognition processes using an image capturing device |
| Automated feedback-based application optimization |
| Preventing software application tampering |
| Reducing database stress using cognitive data caching |
| Video file modification |
| Query processing and access control in a blockchain network |
| Converting whiteboard images to personalized wireframes |
| Application testing on a blockchain |
| Mobile telephone number selector |
Consulting
Fixed-scope engagements
| Service | Price |
|---|---|
| AI Strategy Review | ₹25k+ |
| Architecture Assessment | ₹40k+ |
| Computer Vision PoC | ₹75k+ |
| Agentic AI Development | ₹1L+ |
First Batch Live Workshop
Live CodingBeginner Friendly
Live CodingBeginner FriendlyBuild an AI-Powered Portfolio Website in 60 Minutes
A live, hands-on Agentic AI session where you turn a CV, projects, skills, and experience into a polished portfolio website.
Outcome
A working portfolio site
Includes
Source code and prompts
After class
Recording for revision
Book via Calendly
Exclusive consultation booking section
Book a dedicated discussion for AI strategy, architecture, or implementation planning.
Open CalendlyContact
Consultation and project discussions
Direct Contact
LinkedIn: jhilam-bera-0ba006b0
GitHub: github.com/Jhilam2014
YouTube: @jhilam3017
WhatsApp: +91 9740488875
Email: jhilam.astro@gmail.com
Phone: +91 9740488875
How Agentic Build Worked this web site
Orchestrator-agent multi-agent workflow
Orchestrator
Parsed goals, decomposed features, created task graph and dependency order.
UI Agent
Built reusable sections, theme-safe styling, and iconography across light and dark modes.
Content Agent
Mapped CV, projects, patents, and consulting details into section-specific narratives.
QA Agent
Ran build checks, found hydration/runtime risks, and shipped stable integration fixes.
Stage 1: Planning
Input requirements, risk checks, and execution plan are created.
Stage 2: Parallel Execution
UI, content, and integration tasks run in parallel with frequent handoffs.
Stage 3: Validation + Ship
Build/test verification gates merge changes into production-ready output.
Detailed Handoff Flow
Extra Curriculum / Hobbies
DIY projects and content
YouTube Channel: youtube.com/@jhilam3017
Featured themes: DIY Astrophotography with Raspberry Pi, Robotics experiments, and AI project walkthroughs.
Milky Way Galaxy
Milky Stacked
A stacked wide-field Milky Way capture showing the dense star field and galactic band structure. The stacked processing helps bring out faint sky detail while reducing noise.
Orion Nebula
Orion
A personal capture of the Orion Nebula region, one of the brightest deep-sky targets visible from Earth. The image highlights the glowing nebular core and surrounding star field.
Moon
Moon
A lunar video captured through my custom-made telescope using a Raspberry Pi ROS robotic system with an RPi HQ camera. The clip shows hands-on astrophotography hardware, robotics control, and real lunar observation in one build.
