Anchit Shrivastava

Assistant Manager & AI Lead (Dec 2025 – Present) · Lead AI Specialist & Architect (Mar 2024 – Dec 2025)

Kuala Lumpur, Malaysia · Hybrid

• Lead all of Joget's AI initiatives — the "Joget Intelligence" suite in Joget DX 9 (AI Designer, AI Composer, AI Agent Builder, AI Assistant, AI Form Fill) — with a lean team of two.
• LLM-powered app generation cut app build time ~70% for citizen developers and tripled the self-service rate.
• Fine-tuned Llama 3, Mixtral 8x7B, and Cohere Command R on distributed GPU clusters (DeepSpeed, NCCL/Gloo, Flash Attention, PEFT); served via vLLM + Celery + Redis on AWS EKS with advanced RAG over live business context.
• Built an adaptive agentic assistant that learns from user feedback via contrastive-loss retraining (Ray, Redis, SageMaker), with a weekly auto-refreshed knowledge base that reduced hallucinations.
• Built the AI Agent Builder on a self-implemented LangChain/LlamaIndex-style orchestration architecture, plus an AI Form Fill Element that populates structured forms from raw text via LLM extraction with guardrails.
• Plugins run in production at Fortune 500 companies; Joget was recognized in Everest Group's Innovation Watch for Generative AI in Software Development alongside Google and Meta.

Remote · Global

• Founded an applied-AI studio serving clients across the US, Europe, Singapore, and India — spanning fintech, healthcare, legal, HR tech, FMCG, and media.
• Built an AI voice front desk for a Houston healthcare practice-management firm that fully automates ~70% of inbound calls with 24/7 booking via intelligent IVR.
• Delivered real-time crypto portfolio intelligence, contract & invoice intelligence (~80% less manual review), and two AI products taken 0→live (hospitality job-matching and a mutual-fund advisor).

Associate Data Scientist (Mar 2023 – Mar 2024) · Junior Data Scientist (Apr 2021 – Mar 2023)

India · Full-time

• Led teams of 5–20 across multiple client engagements as project scope grew.
• Led ChatFactory for PathFactory (client) end to end — a production RAG system transforming B2B content consumption with funnel-stage accelerators.
• Built a News & Journalism Analysis platform for a government initiative — 300K+ articles from 160+ Indian news sources with custom NER, sentiment, and topic modeling (GuidedLDA, BERTopic); mapped 10K+ authors and reached out to 6K+.
• Trained decoder-only LLMs from scratch on multi-GPU DeepSpeed setups and advised clients on LLM costing and use-cases — pre-dating the AI leadership role at Joget.
• Delivered computer-vision influencer discovery for a global F&B corporation (fine-tuned ResNet50/VGG19; 500+ qualified micro-influencers per region) and B2B account intelligence with SpaCy/SPERT extraction and NetworkX knowledge graphs.

Mentor (May 2021 – Jul 2024) · Data Scientist Intern (Jul 2020 – Dec 2020)

India · Part-time

• Applied deep learning (TensorFlow) to data-for-social-impact problems and mentored aspiring data scientists across a four-year involvement.
• Ran political-discourse and misinformation analyses whose findings were cited by Al Jazeera in reporting on Indian politics and media.
• Mentored a student team building an Indian Sign Language translation system (YOLO/SSD, OpenCV) — extending the accessibility focus of my granted patent for deaf and hard-of-hearing users.

Cited in:Al Jazeera — Reclaiming Bollywood, Shah Rukh Khan styleAl Jazeera — India, a hotbed for Israel–Hamas war misinformationThe Reporters' Collective — Electoral bonds investigation