Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

AI engineer focused on applied AI systems across healthcare and fintech platforms. Experience spans production deployment, cost and latency optimization, and building retrieval-augmented and compliance-focused AI assistants. Worked on hybrid RAG systems, data integrity diagnosis in large-scale pipelines, and Kubernetes/Azure migrations. Also supported legal decision classification using transformer-based models and contributed to product compliance solutions with role-gated assistants and automated workflows.…AI engineer focused on applied AI systems across healthcare and fintech platforms. Experience spans production deployment, cost and latency optimization, and building retrieval-augmented and compliance-focused AI assistants. Worked on hybrid RAG systems, data integrity diagnosis in large-scale pipelines, and Kubernetes/Azure migrations. Also supported legal decision classification using transformer-based models and contributed to product compliance solutions with role-gated assistants and automated workflows.

Lakshit Asoni

AI Engineer, Data Scientist, DevOps Developer





AI engineer focused on applied AI systems across healthcare and fintech platforms. Experience spans production deployment, cost and latency optimization, and building retrieval-augmented and compliance-focused AI assistants. Worked on hybrid RAG systems, data integrity diagnosis in large-scale pipelines, and Kubernetes/Azure migrations. Also supported legal decision classification using transformer-based models and contributed to product compliance solutions with role-gated assistants and automated workflows.…AI engineer focused on applied AI systems across healthcare and fintech platforms. Experience spans production deployment, cost and latency optimization, and building retrieval-augmented and compliance-focused AI assistants. Worked on hybrid RAG systems, data integrity diagnosis in large-scale pipelines, and Kubernetes/Azure migrations. Also supported legal decision classification using transformer-based models and contributed to product compliance solutions with role-gated assistants and automated workflows.

Available to hire

AI engineer focused on applied AI systems across healthcare and fintech platforms. Experience spans production deployment, cost and latency optimization, and building retrieval-augmented and compliance-focused AI assistants.

Worked on hybrid RAG systems, data integrity diagnosis in large-scale pipelines, and Kubernetes/Azure migrations. Also supported legal decision classification using transformer-based models and contributed to product compliance solutions with role-gated assistants and automated workflows.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Intermediate

Intermediate

Intermediate

Intermediate

Intermediate

Intermediate

Intermediate

Intermediate

Language

Amharic

Intermediate

Work Experience

Intern, Applied AI Systems at Flipkart Internet Pvt Ltd

June 1, 2026 - August 31, 2026

Reduced Azure egress bandwidth costs (~99.8%) and improved end-to-end latency (~400ms to ~50ms) by designing a caching and locking strategy across a Kubernetes cluster (150+ nodes, 3,400+ pods) handling 3,000,000+ daily runs. Eliminated redundant image rewrites and stale-image pulls. Built a price-comparison pipeline for the Pricing team by processing authenticated vs unauthenticated pricing data across 30 zones and 1000+ SKUs. Diagnosed data integrity faults behind wrong results including an unfiltered scan on an 810M-row table, a zip-code type mismatch, and silently pruned source data. Improved match coverage from 17 records to 56,873 (66%). Migrated a 150+ VM fleet handling 3,000,000+ daily jobs from Debian 11 to Debian 13 by resolving Kubernetes package signature issues, relocating container runtime repositories, and storage-driven pod eviction improvements (node disk usage 99% to ~70%).

Intern - Applied AI Systems / Engineering at Flipkart Internet Pvt Ltd

June 1, 2026 - August 1, 2026

Reduced Azure egress bandwidth costs by ~99.8% (~$52,200/mo to $120/mo) and improved end-to-end latency from ~400ms to ~50ms by designing cache and locking across a Kubernetes cluster (150+ nodes, 3,400+ pods) running 3,000,000+ daily runs. Eliminated redundant image rewrites and stale image pulls. Built a price-comparison pipeline for the Pricing team to join authenticated vs unauthenticated price data across 30 zones and 1000+ SKUs. Diagnosed data-integrity faults behind wrong results, including an unfiltered scan on an 810M-row table, a zip-code type mismatch, and silently pruned source tables; increased match coverage from 17 records to 56,873 (66%). Migrated a 150+ VM fleet handling 3,000,000+ daily jobs from Debian 11 to Debian 13 by resolving Kubernetes package signature issues, relocating container runtime repositories, and storage-driven pod eviction changes (node disk usage 99% -> ~70%).

Founding AI Engineer at VeLyra Labs

May 1, 2025 - Present

Sole engineer on a production compliance suite for tax & GST: built 8 role-gated assistants for law/rule comparison, TDS computation, notice validation, and draft generation. Worked with 10+ practicing CAs, conducted firm discussions for organization-wide rollout, and led engineering onboarding for new engineers. Built a Tax Research Tool: a hybrid RAG (BM25 + text-embedding-3-large with RRF fusion, parent/child chunking, domain-boosted reranking) with streamed answers, in-line citations, and warnings when a provision has been superseded by future-dated amendments. Implemented cost safety using Azure Cost Management triggering a Logic App that flips a runtime flag at 100% spend to disable the product before budget overrun (no human-in-the-loop required). Worked on healthcare platform services (EHR, scheduling, inventory) integrating real HL7 FHIR and DICOM via Azure Health Data Services, with Auto Agents running clinical research over ClinicalTrials.gov and OpenFDA.

Machine Learning Intern at Legals Log

September 1, 2024 - February 1, 2025

Classified 10,000 legal judgments into Facts, Issues, and Conclusions using Legal-BERT and RoBERTa, improving true positive rate by 40 percentage points over baseline (from 69%). Built clustering pipelines (K-Means, DBSCAN, Agglomerative) for semantic grouping across 250K+ legal texts to enable structured discovery across previously unorganized case archives.