Commercial Experience
Professional background aligned with Deepak Yadav on LinkedIn. Use the + controls in work history to open full role detail.
Professional profile
Deepak Yadav
Lead Product Research & Development Engineer
Lead Product Research & Development Engineer, bringing 20+ years of IT experience and graduate training in software design and artificial intelligence to market-research and consumer-intelligence products. Current work centers on Agentic AI and LLM systems—production LangGraph / LangChain supervisors with specialized agents, RAG with vector retrieval, MCP-style tooling, and observability—plus CrewAI-orchestrated LLM pipelines for extraction and research workflows. Core ML and NLP remain central: unsupervised and supervised learning on text, multi-level classification, deep models (including LSTM gains from ~85% to ~98% on a high-stakes binary task), and OpenCV-based imaging, alongside earlier delivery in cloud migration, microservices, Spring ecosystems, and Spark-based platform work.
Total experience: 20 + years
View full profile on LinkedInAt a glance
- 20+ years of IT experience (including last 6+ years in AI specially) across many countries India (7.5), Singapore (2.5), US(1), UK(1) & Ireland(9).
- Technological forte is Python / AI / ML-Ops / Agentic Framework / RAG / LLM with Azure & AWS.
- My Technical Qualification (Bachelor of Technology in Computer Science & Eng.) as well as MS (in AI & Software Design) together with Post Graduation Diploma in IT Systems Mgmt. (equivalent to MBA) gives me firm foundation in understanding requirements from business point of view and finding their solid technical solutions.
- Being strong believer of "Lifelong learning", I completed MS in Artificial Intelligence from TUS (Technological University of Shannon), Athlone, Westmeath in 2021.
- Worked with leading global brands in several domains such as Investment Banking (UBS-AG Swiss Investment Bank), Retail Banking (GE Money SBI of India), Retail (Walgreens Co. & BestBuy of USA and Screwfix of UK) and Travel & Shipping (Maersk Line), LM Ericsson (Telecom), Fidelity & AIB in Ireland.
- Worked with India's top 3 IT companies like Infosys (3 years), TATA Consultancy Services (4 years) and HCL Technologies (2 Years).
- Avid traveler to 20+ countries, fostering adaptability and cultural fluency.
Work experience
Each role shows a summary line; click the + on the right to open full project and responsibility detail.
-
Oct 2021 – Feb 2026 · 4 yrs 5 mos
Sr. Machine Learning Engineer(Sr. Manager) — Circana · Ireland Show or hide full role description
- Production LangGraph + LLM Agentic framework: Led design and implementation of a production-scale LangGraph (on LangChain) + LLM Agentic stack: hierarchical supervisor, specialized helper agents, RAG with Weaviate, MCP for extensible tool use, and AgentOps—circuit breakers, failover, and health checks—so multi-agent customer operations stay reliable under planning and ReAct-style flows while limiting impact when individual agents fail. Outcome: quantify social-media influence (authenticity, emotional resonance, temporal dynamics, causality), fully automate long-form reporting, and eliminate the manual effort of a five-person analyst team. System architecture and production deployment mirror the Agentic production reference stack (supervisors, RAG, MCP, Kubernetes-style hardening).
- CrewAI + LLM product extraction: Spearheaded an LLM-driven product information extraction system for complex HTML product pages—fine-tuned open-source LLMs plus RAG, orchestrated through a CrewAI Agentic framework—significantly improving extraction accuracy and speeding competitive intelligence and market research pipelines.
- Core ML & NLP: Unsupervised learning with multiple clustering approaches on large text-based consumer and survey datasets.
- Text classification: Multi-level waterfall classification pipelines using NLTK and complementary libraries across heterogeneous user-generated text.
- Deep learning: Improved a high-stakes binary classification task from ~85% to ~98% accuracy using LSTM-based neural networks.
- Computer vision: Image processing and analysis with OpenCV, alongside text models where workflows combine visual and textual signals.
- Generative AI & modern NLP (research collaboration): Industry-side collaboration with academic research (TUS) on consumer communications—ensemble methods combining generative AI with LSTM for email classification, receipt detection, and misclassification reduction using real Circana-sourced data.
- Intent & truncated text: Work aligned with doctoral supervision on intent classification in truncated consumer messages, including comparative study of classical models (Naïve Bayes–Bernoulli, SVM, Random Forest with bag-of-words) and deep models, plus evaluation of GPT-2 / GPT-Neo-class language models for short, incomplete user text—relevant to chatbots, routing, and automated understanding pipelines.
- Research leadership: External supervisor on published doctoral research documenting the above (see Research & publication below)—bridging production ML engineering with rigorous evaluation and explainable ensemble design.
-
Jan 2019 – Oct 2021 · 2 yrs 9 mos
Consultant Contractor, Sr. Developer — Cloud & AI — Samdiv Technologies Ltd. · Ireland Show or hide full role description
- MLflow + data lake (greenfield AIB): Built MLflow and data lake infrastructure from scratch for a greenfield AIB AI/ML programme—enabling experiment tracking, model versioning, and scalable data management across AI workflows.
- Django AI microservices (AIB POCs): Developed multiple Django-based AI microservices as POCs for AIB Bank, demonstrating production-minded LLM and ML capabilities for banking use cases.
- Implemented alpha trades classification end to end using Python ML implementations on AWS.
- Migrated CI/CD from Jenkins–Chef to Jenkins–Ansible for repeatable, playbook-driven deployments.
- Built new Spring Boot microservices (REST/API-style services) for Open Banking requirements—clear service contracts and testable boundaries.
- Monolith decomposition: Re-architected and split a monolith into microservices, then migrated services to AWS following the target reference architecture (compute, networking, and supporting platform services as per the plan).
- Cloud build-out: Implemented cloud-side components and supporting services the migration depended on—aligned to the architecture blueprint and operational needs.
-
Oct 2016 – Dec 2018 · 2 yrs 2 mos
Consultant Sr. Developer — Sidero / Ericsson · Athlone, Ireland Show or hide full role description
- AWS & microservices migration: Migrated a monolithic application toward the AWS cloud and a microservices-style decomposition—defining service boundaries, migration sequencing, and cloud-hosted runtime targets.
- JVM performance: Runtime performance evaluation across multiple parameters—especially heap memory and related JVM settings—plus application-level fixes where profiling pointed to bottlenecks.
- Spark rewrite: Rewrote computation logic using the Spark API where batch/analytics-style workloads outgrew single-node processing—improving throughput and maintainability versus legacy in-process logic.
- Enterprise integration: Continued integration-heavy delivery in a telco / Ericsson consulting context—hardening services for operations, clarifying API and data contracts, and supporting handover to client maintainers.
-
Feb 2015 – Oct 2016 · 1 yr 8 mos
Technical Architect & Sr. Application Developer — HCL Technologies · Noida Show or hide full role description
- Time-critical POCs: Single-handed delivery of multiple proofs of concept introducing first-time tools and techniques into live programmes, on very short timelines—de-risking adoption before wider rollout.
- Architecture translation: Partnered with Solutions Architects and Engineering so high-level solutions were correctly reflected in Low-Level Designs that teams could actually ship.
- OAuth 2.0 (Digi): End-to-end OAuth 2.0 on Digi web properties—flows, token handling, integration touchpoints—and ownership of related defects until stable in production.
- Design depth: Authored Mid-Level and Low-Level technical specifications, integration contracts, and phased delivery plans under programme governance.
- Stakeholder & risk: Bridged enterprise architecture and sprint delivery—making scope, cross-team dependencies, and technical risk visible while designs were still open, so stakeholders could weigh trade-offs in time and avoid late, expensive rework.
-
Nov 2010 – Jan 2015 4 yrs · 3 assignmentsTata Consultancy Services
One employer, successive client projects — each with its own dates, location, and role title.
Apr 2014 – Jan 2015 · 9 mos Technical Architect & Sr. Application Developer · Gurgaon, India Show or hide Gurgaon assignment details
- Architecture hand-off: Worked with Solutions Architects and Engineering so high-level programme solutions were translated into Low-Level Designs that the engineering teams could build and test.
- Technical analysis, tracking, and estimation for scope changes, dependencies, and release planning.
- Enterprise integration: Integrated applications running on heterogeneous platforms using advanced Spring Integration patterns—message channels, adapters, transformers, and orchestrated flows between legacy and modern tiers.
- Design Artefacts: Produced Mid-Level and Low-Level technical plans, service boundaries, and interface contracts so squads could implement consistently across modules.
- Engineering alignment: Partnered with delivery teams through build, integration, and defect cycles so design intent matched what shipped to production.
Jun 2013 – Mar 2014 · 9 mos Technical Architect & Sr. Application Developer · Yeovil, UK Show or hide Yeovil assignment details
- Architecture hand-off: Partnered with Solutions Architects and Engineering so programme-level solutions were translated into Low-Level Designs that engineering squads could implement.
- Technical analysis, tracking, and estimation across work packages and releases.
- Enterprise integration: Connected applications on heterogeneous platforms using advanced Spring Integration (and related Spring-stack patterns) for reliable messaging, adapters, and cross-system workflows.
- Design Artefacts: Authored Mid-Level and Low-Level technical plans, interface contracts, and implementation guidance aligned to the client’s delivery governance.
- End-to-end ownership: Integration-heavy delivery with the UK client engineering team—design coherence through build, integration test, and defect cycles in a distributed model.
Nov 2010 – May 2013 · 2 yrs 6 mos Tech Lead & Sr. Application Developer · USA (Minnesota) and India (Delhi) Show or hide USA and India assignment details
- Sprint execution: Sprint planning, user-story and task breakdown, and assignment in JIRA with clear ownership.
- Definition of done: Established acceptance criteria (“steps of doneness”) so sprint closures reflected real completeness, not partial delivery.
- Quality bar: Monitored adherence to Test-Driven Development (TDD) and encouraged test-first habits across the team.
- Hands-on engineering: Wrote application code for larger or higher-risk backlog items to meet quality standards and deadlines when complexity demanded it.
- Technical walkthroughs & code reviews: Reviewed designs and merges for correctness, regressions, and maintainability; used reviews to spread patterns and reduce rework.
- Cross-site delivery: Coordinated engineering outcomes across onshore (US) and offshore (Delhi) squads—alignment on priorities, dependencies, and integration checkpoints.
-
Nov 2008 – Sep 2010 · 1 yr 10 mos
Sr. Software Engineer — Birlasoft Show or hide full role description
- Requirements & legacy systems: Analysed requirements and built a solid understanding of the existing system; maintained clarification logs to remove ambiguity between on-site and offshore teams.
- Change impact: Assessed impact of each change before coding to control regression and scope creep.
- Implementation: Design-to-code delivery across assigned modules.
- Unit testing: Generated unit test cases, executed module-level unit tests, and refreshed cases as behaviour evolved.
- Standards & reviews: Enforced coding standards and participated in reviews (code / deliverable) before handoff.
- System testing & defects: System testing cycles; defect log preparation, resolution, retest, and closure with stakeholders.
-
Aug 2005 – Sep 2008 · 3 yrs 1 mo
Software Engineer — Infosys · Singapore Show or hide full role description
- Applications: Converted algorithmic specifications into code; modified implementations to match design; incorporated ongoing change requests safely.
- Analysis & integration: Analysed programs to recover missing behaviour; probed prototype modules; integrated subsystems into the evolving design with milestone-based verification.
- Testing & quality: Ensured completeness of testing for assigned features before handoff to downstream stages.
- Client interface: Worked with the client to capture evolving requirements and reflect them in technical tasks.
- Database: Database-related analysis and implementation in step with application releases—SQL/data-access changes, integrity constraints, and tuning where performance or correctness required it.
-
Jun 2005 – Jul 2005 · 1 mo
Software Developer — Indiagames Ltd. · Mumbai Area, India Show or hide full role description
- Key member of the working-model development team; R&D-style coding.
- Integrated software from different vendors to add multiplayer features to standalone games.
- Troubleshooting to improve code cohesion.
- Testing for runtime memory requirement and efficiency; code changes to reduce memory use and maximise efficiency on device constraints.
-
Dec 2004 – May 2005 · 5 mos
Developer — VRG Electronics · New Delhi Area, India Show or hide full role description
- R&D-oriented coding; ported games to multiple handsets (Nokia, Sony Ericsson, Motorola, and others).
- Bytecode-level JVM troubleshooting during porting (manual analysis and tools: Mobile JBuilder, NetBeans, Sun ONE Studio v4).
- Testing for memory and efficiency; optimisations to reduce runtime memory and improve performance on constrained devices.
Education
- MSc (Honors), Software Design and Artificial Intelligence — Athlone Institute of Technology (TUS Athlone), Ireland · 2019 – 2021 · 2 yrs
- Post Graduate Diploma, IT and Systems (IT Management) — All India Management Association · 2006 – 2008 · 2 yrs
- B.Tech, Computer Science — Ideal Institute of Technology · 2000 – 2004 · 4 yrs
Additional background: secondary schooling (Universal Public School) and formal course work including MBA (IT & Systems).
Licenses & certifications
- ITIL V3 Foundation — BIS · issued Jan 2012
- Advance Certificate in Finance (investment banking, portfolio management, risk management) — All India Management Association · issued Jun 2008
Research & publication
Doctoral research at Technological University of the Shannon credits Deepak Yadav as external supervisor on work combining generative AI and LSTM ensembles for email and intent classification, with Circana collaboration for real consumer data.
A generative AI enabled ensemble machine learning strategy for intent classification in truncated consumer communication
View on TUS ResearchLanguages
English