Anmol Jain

Client: Orchid Soft Solutions (surrogacy & fertility technology). Built two production AI platforms for clinical-record summarization and contract-driven expense compliance.

AI Engineer at Telomere (2025-08 – 2026-04)

Client: Orchid Soft Solutions (surrogacy & fertility technology). Built two production AI platforms for clinical-record summarization and contract-driven expense compliance.

• Gather AI: Multi-Agent Clinical Records Summarization Platform
• Built an end-to-end platform (NestJS, React 19) that converts hundreds of pages of fragmented obstetric records into structured, citation-backed summaries, reducing chart review from hours to minutes.
• Designed a multi-agent pipeline with specialized agents for pregnancy timelines, family history, prenatal care, delivery events, vaccinations, and clinical risks, followed by consolidation and verification stages.
• Engineered document processing using AWS Textract, BullMQ/Redis, and chunked LLM extraction across 10M+ character records, with page-level citations for every extracted clinical fact.
• Built structured-output workflows on Gemini 3 Pro with a Claude-based evaluation agent, deterministic clinical safeguards, and deployment through AWS CDK, ALB auto-scaling, and OpenTelemetry.
• SeedTrust: Agentic DR Request & Contract Compliance Platform
• Built a contract-aware RAG system (FastAPI, LangChain, PostgreSQL/pgvector) that validates surrogate disbursement requests against 50+ page agreements and grounds every decision in cited clauses.
• Designed an agentic decision pipeline combining request extraction, multi-query retrieval, contract reasoning, tenant filtering, and decision verification to automate multi-day manual reviews.
• Implemented a deterministic self-consistency verifier that overrides decisions when the model's outcome contradicts its explanation or cited evidence.

AI Engineer at Accacia (2025-01 – 2025-08)

• Designed and deployed an AI-driven outreach platform for hyper-personalized B2B messaging (FastAPI, Supabase, PostgreSQL), onboarding 10K+ users with analytics through PostHog.
• Built multi-agent pipelines for lead research, tone control, and qualification, using entity extraction and semantic scoring across company, industry, and buyer-persona context.
• Engineered an AI-powered Chrome extension automating LinkedIn and Gmail outreach through real-time, context-aware message generation using OpenAI and Perplexity.
• Built an AI pitch-deck generation system using Neo4j and GraphRAG to model company, industry, and persona relationships and generate personalized PowerPoint sales decks.

Software Engineer at Turtlemint (2022-08 – 2024-10)

• Architected an advanced document-understanding system using LayoutLM, achieving a 95% F1-score across 103 entity classes and 5,000+ document formats.
• Developed an LLM extraction pipeline using Claude Haiku, achieving 98% accuracy across documents from 40+ insurers.
• Engineered a configurable framework for prompts, extraction schemas, and key types, supporting the processing of 10M+ policy documents.
• Built a custom PDF parser using PyMuPDF and pdfminer, achieving 99% parsing accuracy across insurance-document domains.
• Led model deployment using TorchServe and AWS SageMaker, supported by a high-throughput architecture using Kafka and RabbitMQ.
• Scaled the platform to handle 10K+ requests per hour with automated retraining, evaluation, and monitoring pipelines.
• Reduced document-processing latency from 11 minutes to 11 seconds through optimized preprocessing, OCR, and inference pipelines.
• Integrated AWS Textract with SVM-based page ranking for intelligent page selection across 1M+ documents.

Tech Intern at Turtlemint (2022-02 – 2022-08)

• Built analytics dashboards using Apache Superset and Metabase, enabling business insights across MySQL and PostgreSQL.
• Developed APIs for evaluating document-data accuracy using OCR confidence scoring, string matching, and configurable validation rules.

BTech in Computer Science – Vishwakarma Institute of Technology (2022-06)