Nishant S.

🏆 Expert-Vetted, Top Rated. Senior LLM/ML/AI Architect and entrepreneur (Ph.D., CMU. IBM Research). 20+ years of hands-on experience shipping frontier AI products. Co-build with your team. Help leadership make sound AI decisions. Avoid costly mistakes and wasted months in AI agent projects. Upwork: 10+ years | 70+ Projects | $500K+ delivered | Production ML across 10+ industries | Top-Rated Plus Focus areas: ✨ architecting reliable Multi-Agent workflows ✨ making internal knowledge base more accessible ✨ build, optimize complex software products with Agents ✨ context, harness optimization ✨ automating repetitive work Recent projects involve: ✨ Agentic search and RAG Agents across company data. ✨ Orchestrate reliable LLM Agent workflows: Claude, Codex, Skills, Langchain, Memory, Worktrees, Harness optimization. ✨ Migrate from costly cloud-based Voice platforms (Vapi, Retell, Bland) to cost-effective on-premise alternatives (Livekit, Pipecat). Optimize Latency and accuracy of ASR + LLMs + TTS pipelines. ✨ Building speech recognition for custom domains - Whisper, pyannote diarization, noisy multi-speakers. Whisper - domain-specific finetune, real-time, speech-text alignment, diarization. ✨ Training deep learning custom models. ✨ Mentoring teams, taking products to market ✨ Coaching: Multi-Agent systems, RAG, LLMs, Voice Agents. Claude, Codex. Many companies treat building workflow Agents as a casual week-long project. This only leads to wasted time and resources. My article titled - Why your GPT + Vector Search demo won't make it to production - (linked in my portfolio) discusses a disciplined approach to building search agents. --- Background -- I finished my Ph.D. from Carnegie Mellon and worked at IBM Research, building AI solutions and distributed systems tools. For almost a decade, I have helped many clients through the entire Machine Learning lifecycle, building wide variety of deep learning models for text, vision and speech problems. My projects include a mixture of creative, algorithmic problem solving and building across the tech stack. I bring a very unique combination of skills to the table: ↪ I’m a senior AI engineer: Strategic problem solving with cutting edge AI tech. - Reliable Multi-Agent Systems (Codex, Claude CLI/APIs, Skills, Memory) - Voice Agents (Pipecat, Livekit, Vapi) - LLM Agents (OpenAI, Claude, Gemini, DSPy, prompt engineering, LLM as judge) - DeepResearch/RAG: Qdrant, ElasticSearch, Vespa, Pinecone, Weaviate - model fine-tuning/inference, LoRA, RLHF (Transformers, GPT, ViT, ASR, CLIP), - compression/optimization (Pytorch, TensorRT, Deepspeed, ONNX) - cloud: Docker/AWS - programming Stack: Python, Pytorch, FastAPI. ↪ As an Architect - I go from ambiguous specs to a clean, modular design of a complex system. Plan a system as a whole before optimizing the parts. Keep design simple. Mentor team. ↪ As a Researcher - I can explain and distill out consumable stuff from the most math dense, cutting-edge academic text. Published internationally over 20 years. Solving complex, multi-disciplinary problems is my core strength. ↪ As a Product Entrepreneur - I deeply care about the product strategy: who buys it, who consumes/uses it. I hustle to bridge the gap between cutting-edge tech and impactful products. I spot, lead and mentor deep tech talents. --- Testimonials -- Please have a look at my portfolio to learn more about my past engagements and feedback. 🏆 Expert-Vetted, Top 1% "Nishant is extremely knowledgeable about NLP and LLM incorporation into a project. He is very thoughtful and considered his responses carefully when answering my questions. He provided valuable insight and advice which will save significant time, effort and expense. His advanced expertise and wide-ranging experience, including academic publications as well as business consulting give him exceptional capability. I recommend Nishant highly. " "Nishant is incredible to work with. He is a fount of machine learning wisdom and knowhow. His involvement with our project was instrumental in taking it to new heights."