Bangalore, India

Om Sharma.

Building real-time AI systems that feel instant.

01
Voice Agents
02
LLM Orchestration
03
Inference Infrastructure

I engineer voice agents, orchestrate LLM workflows, and build the infrastructure that makes models feel alive inside two seconds.

View Work→View Resume↗

§ 1 — Introduction

I'm an engineer at Jobtwine, where our team builds real-time voice AI for enterprise interview screening — the kind of system where latency, reliability, and conversation quality decide whether the product works. Earlier I worked on RAG pipelines and LLM fine-tuning at Darwix AI and VDOIT Technologies.

On my own time I build small tools that scratch real itches — JiraGenie, a natural-language CLI for Jira with a self-healing workflow engine, is the current one. I write about what I learn shipping models into production: the gaps where the model second-guesses itself, the network hops that quietly cost you a turn.

§ 2 — Experience

Where I've shipped.

Jobtwine

Jul 2025 — Present

Associate Software Engineer · Bangalore

Built a real-time AI interviewer used by several big firms.

<2s latency · 300+ interviews/day · 3 enterprise clients

View detailsHide details

→Real-time AI interviewer — a streaming STT → LLM → TTS pipeline at sub-2s p50 latency.
→Twilio ↔ LiveKit orchestration (WebRTC / SIP) with token-level TTS streaming cut per-turn latency ~3s.
→Queue-based dispatch with multi-LLM failover, serving several big firms.

Darwix AI

May 2025 — Jul 2025

Software Engineer, AI Systems · Gurugram

Built real-time sales-call analysis and document-ingestion systems.

+40% chunking accuracy · real-time call analysis · 2-channel desktop client

View detailsHide details

→Document-ingestion engine (Pinecone + PostgreSQL) with auto type-detection — +40% chunking accuracy.
→Real-time sales-call analysis: speaker diarization, live transcription & performance scoring.
→Cross-platform Windows / Linux client streaming dual-channel call audio over WebSocket.

VDOIT Technologies

Jan 2024 — Apr 2025

Software Engineer, AI / ML · Gurugram

Built high-scale backends and RAG pipelines for domain-specific LLMs.

100K+ concurrent users · RAG pipelines · STAR performer

View detailsHide details

→Backend for 100K+ concurrent users on Django with a multithreaded architecture.
→Retrieval-augmented generation (RAG) pipelines over vector search for contextual document retrieval.
→Fine-tuned domain LLMs with LoRA / QLoRA. Awarded STAR Performer.

§ 3 — Capabilities

What I build with.

01LanguagesThe everyday tools.8 tools

PythonC++CJavaSQLTypeScriptJavaScriptBash

02AI / MLModels from first principles to fine-tune.14 tools

PyTorchTensorFlowJAXKerasHugging FaceTransformersDiffusersscikit-learnXGBoostNumPyPandasOpenCVCNNsLoRA / QLoRA

03LLM & AgentsOrchestrating models into systems that act.17 tools

OpenAIAnthropic ClaudeGeminiLlamaMistralLangChainLangGraphLangflowLlamaIndexHaystackCrewAIAutoGenPydantic AIMCPRAGFunction callingPrompt engineering

04Realtime & VoiceConversations that answer in under two seconds.12 tools

LiveKitTwilioPipecatDeepgramCartesiaElevenLabsWhisperSilero VADWebRTCSIPWebSocketsStreaming STT / TTS

05Inference & ServingMaking models fast and cheap in production.11 tools

vLLMTensorRTCUDATritonTGIllama.cppOllamaONNXRayModalQuantization

06Retrieval & DataSearch, memory, and the stores behind them.13 tools

PineconeWeaviateQdrantChromaFAISSPostgreSQLMySQLMongoDBRedisElasticsearchKafkaAirflowSpark

07BackendAPIs and services that hold up under load.10 tools

FastAPIDjangoFlaskNode.jsExpressRESTgRPCGraphQLCeleryPydantic

08InfrastructureWhere it all runs.11 tools

DockerKubernetesAWSGCPVercelLinuxNGINXTerraformHelmCI / CDGitHub Actions

09ObservabilitySeeing what the system is actually doing.8 tools

PrometheusGrafanaSentryOpenTelemetryLangSmithLangfuseWeights & BiasesMLflow

10AI Dev ToolsHow the work gets shipped, faster.5 tools

Claude CodeCursorWindsurfGitHub Copilotv0

§ 4 — Certifications

What I've earned.

Anthropic
AI Fluency: Framework & Foundations
Issued Jul 2026 · ID etenep6o89iu
Verify

§ 5 — Personal projects

What I've built.

Now · June 2026

Speculative
decoding,
in progress.

Building a toy implementation of the Leviathan et al. paper — draft model + target model + measured speedup. Companion blog post in July.

2026 · Open source

JiraGenie

Natural-language CLI for Jira. LLM intent parser. Self-healing workflow transition engine — walks workflow graphs with a loop guard, auto-repairs failed transitions.

2024 · PyTorch